Connecting people, devices, and systems, M2M magazine is dedicated to helping vertical industry thought leaders realize the value of machine-to-machine communications through remote monitoring, RFID, sensor networking, smart services, telematics, and telemetry: the Six Pillars of M2M.

If You Ask It, They Will Answer

19 Industry Leaders Lend Their Insight on What Issues You Should Consider When Deploying a Sensor Network

There are a number of topics that can stir up quite the discussion among those in the WSN (wireless sensor networking) world: what the “right” networking protocol is, how to best power sensor nodes, or even what’s the best way to determine return on investment. To really delve into these topics, Managing Editor Cassie del Pilar and Editorial Director Peggy Smedley brought together 19 companies that are leading WSN innovation and received their insight into the most important business and technological issues potenial WSN adopters need to consider.

But if there’s one thing sensor networking technology providers can agree on it’s this: Sensors are past their experimental phase, and indeed, the technology does really work.

“Sure, there will always be skepticism over the viability of wireless technology for sensor networks: questions of coexistence, stability, scalability, robustness,” says Jimi Simpson, product marketing manager, Jennic Ltd., www.jennic.com, Sheffield, U.K. “But the technology works and the needs exist, and the first movers with the technology are already experiencing the benefits it has to offer.”

To that end, the possible uses for sensor networking are ever increasing, and the M2M industry is working to increase awareness among corporate adopters about what sensor networking technology can offer them. Yet, as sensor networking providers make the business case for the technology, those not well versed with sensor networking are saying, “Yes, we see the advantages, but what does it take to actually implement and maintain these networks?” And their business and technological questions can form quite a list.

Below include a complete collection of the vendors' responses to the 20 most commonly asked questions about wireless sensor networking.

WHAT TECHNOLOGY PROVIDERS ARE SAYING
M2M magazine invited leaders in the wireless sensor networking space to share their expertise on both the business and technological aspects of implementing sensor networking technology.
Accsense Inc. Matt Cheresh, CEO	GreenPeak Technologies Cees Link, CEO
Arch Rock Corp. Brian Bohlig, Vice President, Marketing	Jennic Ltd. Jimi Simpson, Product Marketing Engineer
Augusta Systems Inc. Patrick Esposito, President and COO	Mesh Systems LLC Richard Baxter, President and CEO
Coronis Systems Inc. John Rouse, Director of Sales, Americas	MeshNetics Alex Leonov, Director of MarCom
Digi Intl. Inc. John Schwartz, Technology Strategist	Millennial Net Inc. Martin Hanssmann, President and CEO
Dust Networks Inc. Robert Shear, Director of Market Development	Pedigree Technologies Alex Warner, President and Founder
Echelon Corp. Steve Nguyen, Director of Corporate Marketing	Sensicast Systems Inc. Gary Ambrosino, CEO
Eka Systems Inc. Srini Krishnamurthy, Vice President, Corporate Development	Sensinode Ltd. Zach Shelby, Cofounder and Head of Research
Ember Corp. Bob Gohn, Vice President of Marketing	Sentilla Corp. David Binetti, Director of Marketing
GE Sensing and Inspection Technologies Ted Furlong, Sensing Technology Leader

PART 1: BUSINESS QUESTIONS
1. Bottomline: Are sensor networks past their experimental phase, and does the technology really work?
2. Cost: What determines the total cost of a typical sensor-network deployment?
3. Time: What usually determines the amount of time required to deploy a sensor network?
4. Scalability: How much does the cost and complexity of a sensor network increase with the number of nodes?
5. Management: What is required to manage a sensor network on a daily basis?
6. Troubleshooting: Who is typically responsible for fixing problems with a sensor network after deployment?
7. Interface: How is sensor data usually presented to the end user?
8. Integration: How can companies integrate sensor data with other business systems?
9. Duration: How should adopters approach a temporary installation compared to a permanent one?
10. ROI: What are the best ways to measure return-on-investment for a sensor network?

PART 2: TECHNOLOGY QUESTIONS
1. Resources: What level of technical skill is required to set up and maintain a sensor network?
2. Setup: How is a sensor network installed and activated?
3. Logistics: What are the range limits and physical limitations of a sensor network?
4. Power: What should an adopter understand about power requirements and battery life?
5. Security: How can an adopter make sure its sensor-network data is secure?
6. Reliability: What steps should an adopter take to ensure reliability of the network?
7. Infrastructure: How can a company leverage its existing network infrastructure when deploying a sensor network?
8. Hardware: What criteria should adopters use for selecting a hardware platform?
9. Intelligence: When should a sensor network have distributed processing capability?
10 (a). Standards: What should adopters understand about the standards associated with sensor networking?
10 (b). More Standards: Which proprietary protocols should adopters avoid?

PART 1: BUSINESS QUESTIONS

1. Bottomline
Are sensor networks past their experimental phase, and does the technology really work?

Accsence:
“No question, yes. In appropriate applications, the technology works, and works quite well.”

Augusta Systems:
“Sensor networks are past the experimental phase, but there is still room for improvement in the way in which enterprises deploy these technologies. Intelligent convergence of data from sensor networks and data from other edge devices is a key enterprise requirement. This converged data must become a seamless part of the enterprise IT (information technology) network.”

Digi Intl.:
“The number of deployed nodes that use either point-to-multipoint or mesh topologies has been growing exponentially for the past few years. With millions of new nodes being deployed by multiple vendors each year and with a solid history of installed nodes, sensor networks have grown to become an integral part of many industries.”

Dust Networks:
“Absolutely. Wireless sensor networks have been successfully deployed in a range of harsh industrial environments including oil and gas, food and beverage, pharmaceutical, wastewater, power generation, steel, chemical, commercial printing, and pulp and paper industries. Self-organizing mesh networks deployed in these applications have delivered over 99.99% reliability, and end users are thrilled with the results.”

Echelon:
“Absolutely. Sensors have always been a part of M2M (device networks). While the form of the physical layer may change, the fundamental value of sensor networks is well proven.”

Eka Systems:
“Sensors and technology to monitor and control sensors have been around for decades. Legacy technologies created the proven foundations of the technology and are quite reliable within the confines of their narrow and isolated application spaces. Sensor networking is a data-intensive paradigm, and a shift towards a massive application of proven concepts and technology to very, very large-scale, short-range wireless networking on a global scale with open standards is occurring. Wireless device networking with ad-hoc routing and self-healing capabilities makes it possible now to cost effectively and reliably network tens–of thousands to millions of devices in a single network. Multiple vendors with compliant and low-cost radios and microcontrollers, plethora of module providers, growing pool of software and application providers, capital commitment from large OEMs (original-equipment manufacturers) and the backing of investment community all make up the ecosystem for a very robust sensor networking market with compelling gains.”

Ember:
“Yes. Wireless sensor networks have moved well beyond the experimental phase, supported by mature, robust, and widely deployed ZigBee standard technology. A variety of products in applications such as advanced metering , home and building automation, and home security are shipping and have been commercially deployed around the world.”

GreenPeak Technologies:
“With the arrival of wireless sensor networks, the widespread real-world adoption of wireless sensor networks has finally started in a wide range of industrial, safety, and business applications. A truly wireless solution that functions using very low power levels (without data cables and without power cables) brings ease of installation and ease of maintenance (no need of changing batteries) to sensor networks. When standard compliance (to the IEEE 802.15.4 standard) is added, three of the major requirements for widespread OEM adoption of wireless sensor networks are met.”

Jennic:
“There is no doubt that sensor networks are beyond the experimental phase, and with an ever-increasing number of network deployments of varying size and complexity, it is safe to conclude that the technology really does work. Sure, there will always be skepticism over the viability of wireless technology for sensor networks, questions of coexistence, stability, scalability, robustness, but the technology works and the needs exist, and the first movers with the technology are already experiencing the benefits it has to offer.”

Mesh Systems:
“Sensor networks are clearly past their experimental phase in certain applications, with hundreds of thousands of end points already deployed and providing significant value to adopters. Yes, the technology works. Other applications are currently at the cusp of commercialization, and we expect announcements of new and innovative sensor-network applications in the coming 18 months.”

Millennial Net:
“Yes, definitely past experimental and yes, it really does work depending on the mesh technology utilized. Currently we see ZigBee and emerging standards as unproven today but very important to the development of the market.”

Pedigree Technologies:
“Yes, but right now the market is utilizing these technologies for very focused applications. Like all technologies, improvements are ongoing. Right now it appears to be a question of integration and market acceptance: “Yes, it works, but where should it go first?” In many instances, volume deployments are a market must. The pervasiveness of, especially wireless, sensor-networking technologies will require deep cooperation of infrastructure and carrier organizations, new middleware, and application software organizations and integrators whose business and licensing models align with the volume interests of the sensor network platform companies. The advent of “new sensor-network applications” that are “asset management” in focus have a strong ability to meet these market requirements.”

Sensinode:
“Thanks to the IEEE 802.15.4 standard, a stable global radio technology has been available since 2003 and really works. With the standardization of 6loWPAN (IPv6 over low-power wireless personal-area networks) in 2007, IPv6 technology is now also available for 802.15.4. This provides sensor network designers with a stable platform based on IP (Internet protocol) standardization and creates the basis for wide deployment. There is still work left to do in enabling large, open multivendor deployments, and IP standardization is a major step in making this possible. Additonally, dealing with the huge amount of potential data produced by M2M (machine-to-machine) services presents a large, but achievable challenge.”

Sentilla:
“Yes, the technology is mature, and companies are making real deployments in the real world, saving real money, and creating real revenue opportunities. The envelope is constantly being pushed, of course, introducing new features and functionality, but the core capabilities of sensor networks are being deployed right now in production environments outside of the lab and in the real world.”

2. Cost
What determines the total cost of a typical sensor-network deployment?

Accsence:
“The technology and product cost is not the largest factor. The largest factor is the learning curve of the customer on what to do with the information now provided by the sensor network, information that was previously unavailable.”

Augusta Systems:
“The costs of a typical sensor network deployment is usually driven by five factors: the network architecture, the number of nodes, the diversity of the nodes, the standards and protocols involved in the network communications, and the volume of enterprise applications that will be empowered to receive data from the sensors. The more that can be done to ensure the sensor network is deployed in a manner that guarantees intelligent convergence of data with other data points and integration with enterprise networks, the lower the total cost of ownership in the long run.”

Coronis Systems:
“The actual hardware tends to be the least expensive component in sensor networks, at least in terms of per-unit cost! One thing that needs to be taken into account is the costs for pilot testing, installation, and later on, battery replacement in many cases. Innovative projects most often include custom software application development as well. However, once sensor networks are up and running, operating costs can be extremely low.”

Digi Intl.:
“The cost of the networks certainly includes the sensor and the other associated hardware, but in addition to these factors, there are costs associated with installation and ongoing monitoring or maintenance. Compared to a network that is cabled, the hardware and installation costs end up being more inexpensive, in many cases, for a wireless network.”

Dust Networks:
“Cost is driven by several factors: equipment cost, engineering services, installation services, and technician time. Service costs quickly overshadow equipment cost in most legacy star-connected systems.”

Echelon:
“The primary cost determinant is the application that uses the sensor network. Sensor networks, in our opinion, serve the control network. So the question becomes, ‘What will you do with the data collected by the sensor network?’ This in turn determines the ultimate cost of the deployment. In some cases, the sensor network is less costly than the control portion, in some cases not. It really depends on your starting point and the function of the application(s).”

Eka Systems:
“Sensor networks are built on foundations of three main components: the ‘first-mile’ wireless mesh device network, the backhaul data transport service, and the backoffice integration to business applications. The ‘first-mile’ network is an infrastructure deployment play with a one-time fixed cost for the network nodes and gateways. The backhaul is a subscribed transport service from a service provider and is a recurring monthly charge based on ‘pay-by-the-byte’ model. The backoffice integration and service launch cost is driven by application engineers on payroll. Network operations cost is an additional recurring monthly cost to manage, maintain, and operate all the components of the network as a system and typically requires dedicated support staff.”

Ember:
“It’s a combination of sensor equipment, the networking platform, any backend systems, and the labor associated with the deployment. Sensor equipment would include sensors themselves and any additional tools required for deployment.”

GE Sensing:
“Some key factors include the number of sensors, the area of coverage, the distance between nodes, and the nature of the obstructions within the communication paths. Besides base hardware, these factors influence redundancy and signal power requirements.”

GreenPeak Technologies:
“The major costs for deploying a sensor network are power and networking cabling and the labor expense for installing those cables. In wireless sensor networks, installation costs are reduced to a minimum: Sensors are simply wall-mounted, turned, on and ready for operation. They autoconfigure to the controller and are fully operational within seconds.
In some applications, wireless sensor networks can be deployed with minimal additional cost, as fully integrated solutions in high volumes can be added to existing appliance hardware without a significant increase in BOM (bill of material). However, in some applications that need mesh networking to gain full network coverage, dedicated infrastructure router nodes could be needed, which will add a burden on installation cost. Existing IT (information technology) infrastructure could be used to compensate for this need for infrastructure.
By using low-power wireless networks that can run off of batteries or energy harvesting, even more costs can be eliminated.”

Mesh Systems:
“Major components of costs include end point (node) cost, the sensors themselves, and gateways. Installation costs are also significant, although typically far less than wired solutions.”

Millennial Net:
“Device Cost – integrated mesh device with application hardware/software
Range – This is the largest factor in hardware deployment costs due to the potential large number of incremental ‘“routers”’ that may be required if range cannot meet sensor-to-sensor separation adequately in practical deployment scenarios and power needs to be installed to these devices.
Battery Life – In most cases the cost of supplying power to the devices will be more expensive in comparison with the cost of batteries (if they have a life of greater than five years).
Ease of Deployment – Self-healing, dynamic, and self-administrating mesh networks are a must to leverage typical nonexpert installer skill sets and avoid costly deployment (and reconfiguration) costs. This often also avoids expensive predeployment site audits and network configuration costs.”

Pedigree Technologies:
“One must look at an end-to-end approach to the application. There are some applications for sensor networks that are ‘plug-n-play’ from end to end and are uniquely architected to take advantage of the ease of sensor network technologies and their deployments in combination with software-as-a-service offerings. These are very application-specific approaches, but are extremely cost effective. With legacy systems, integration costs combined with application requirements that provide information about the critical nature of the application and deployment size can help determine the cost of a sensor network deployment. Obviously larger networks will require more hardware, leading to higher costs. Still, the powerful combination of cost-effective hardware and the ad-hoc nature of wireless deployments provide a significant cost savings over traditional conduit-run deployments.”

Sensicast Systems:
“The total cost of ownership of a sensor network comprises four elements: product (hardware and software), pre-installation evaluation, system installation, and post-deployment maintenance. Total cost will vary greatly depending whether the network is: (1) a comprehensive, end-to-end system of preintegrated, elements from a single provider, or (2) a network comprised of separate components from different vendors.
In the first case, installation of the network is simple and uncomplicated maintenance is automatic, integration, and operational issues are rare. On the other hand, when sensors, routers, and other networking equipment are sourced from a variety of different suppliers, the cost to integrate, deploy, operate, and manage the network will be higher. Issues caused by interoperability problems (between incompatible components) can also add to cost.”

Sentilla:
“The majority of the cost is represented in the development, installation, and maintenance time required to create and support the application. The costs of the hardware itself are already low and dropping still further the key metric is total cost of ownership, which is highly dependent on initial development (mostly software) and flexibility (the ability of your application to adapt without needing to redeploy the entire system.)”

3. Time
What usually determines the amount of time required to deploy a sensor network?

Accsence:
“In our case, it is the time to establish successful wireless communication, since any poor link in the system can bring the entire system down, or a poor link prevents one node from communicating, and ANY non-communicating nodes implies lack –of success in meeting customer expectations.”

Augusta Systems:
“Deployment time for sensor networks is usually driven by the complexity of the architecture and the integration process. Enterprises often overlook the importance of convergence and integration. Putting a lot of sensors in the field only yields a lot of data. An enterprise must be prepared to do something with that data. There must be sufficient planning in order to intelligently process that data on a distributed basis and integrate it into enterprise networks in an intelligent manner.”

Coronis Systems:
“The amount of time to deploy a sensor network really depends on the size of the network. The network could have five end points or up to over 50,000 end points. A certain amount of time is required for each unit installed in the field, which is why appropriate setup software/hardware tools are essential. Project management of the deployment may be required if the network is very large or installed in multiple areas.”

Digi Intl.:
“The type of architecture, the technical competence of the installer(s), and the number of nodes that have to be deployed are the main determining factors. Also, the availability of deployment or network commissioning software and tools for the installer(s) plays a role as well.”

Dust Networks:
“Network type and architecture is the primary driver of deployment time. Star-connected networks typically require site surveys, careful device placement, remote and/or directional antennas, and significant network tuning. Self-organizing networks on the other hand can be installed without RF (radio frequency) site surveys and with minimal custom antennas or post-installation tuning. As a result, they are installed in a matter of hours instead of several days or weeks required for alternate approaches.”

Eka Systems:
“Commissioning large sensor networks in commercial and industrial environments with tens –of thousands of devices is an involved process and requires detailed planning, site surveys, pilots, backhaul, and backoffice application integration, testing, data validation, launch, and finally large scale rollout. The networking technology hardware and software is factory integrated in the target device, self-configures when powered on, automatically becomes a node in the network and is not a gating factor in the total amount of time required to deploy a sensor network. Simple home networks, however, can be deployed very quickly with plug-n-play devices in a matter of hours.”

Ember:
“The amount of time is a function of the scale and complexity of the network. Smaller sensor networks, such as in home-monitoring and security applications, can often be quickly installed by the consumer in minutes. However, systems in applications such as building automation and high-end home automation may require professional installation. In any case, ZigBee wireless technology is making installations easier and easier.”

GE Sensing:
“Complexity of the site where the sensors are being installed is the primary driver. If the site is predominantly static, it is easy to install. If there is a lot of material movement expected, the network needs to be made more reliable which takes time to test and validate. The type of wireless network (mesh/star) also plays a role on the install time. Finally, the level of training and availability of tools (for radio strength or interference measurement) can also play a role.

GreenPeak Technologies:
“Since wireless sensor networks are designed for easy installation, a node will automatically link to the wireless network once it is powered. This ‘plug and play’ approach enables every technician to execute a 100% correct installation without the need to understand technical aspects of setting up a wireless network. The most important part of the deployment time will be assigned to the integration of the sensor network data with existing IT (information technology) infrastructure.”

Jennic:
“Deployment time comes down to many aspects. The placement of nodes will typically be a relatively quick process as positions are predetermined by the source of data that they will collect. However, site surveying may be necessary to ensure the optimal placement of routers through simple signal-strength and packet-error-rate checking. Where some nodes are line-powered, then wiring will be required this may or may not be quick exercise. Network visualization and performance monitoring are necessary to ensure that the network is optimally configured the latter will generally have to run for a period of time to show results that are representative of the real network for performance measurement to be made. Ultimately, all wireless sensor networks report data and information into a backend system that may need configuring and may even need some software development.”

Mesh Systems:
“In our experience, the biggest determinant of time is the decision-making process to actually move forward with commercialization and deployment of a sensor-network application. Prudent sensor-network adopters will carefully study the true cost and benefit of a solution and need to make a clear and compelling business case. Once a move-forward decision has been made, the technology and component vendors have made it fairly easy to build robust and cost-effective solutions in relatively short order.”

Millennial Net:
“If the network is battery powered (or utilizing power harvesting) and is truly self configuring … then deployment is as straight forward as placing the sensor in the appropriate location and confirming sufficient signal strength and mesh redundancy.”

Pedigree Technologies:
“The biggest determining factor for deployment times typically has to do with the intelligence and scalability of the network software stack, the wireless transmission environment, plus the number of nodes required.”

4. Scalability
How much does the cost and complexity of a sensor network increase with the number of nodes?

Accsence:
“Greatly, but mostly from the user standpoint in managing the information now possible.”

Arch Rock:
“When starting with a scaleable architecture and proven deployment, management, and integration tools, the cost and complexity can be managed and minimized. Intelligent architectural choice upfront determines ease of scalability in the future as the network grows. By choosing stateless networking egress points, customers benefit from near-linear throughput gains, and reduced latency within the mesh. The right aggregation and routing technology can turn a large sensor network into a repeating series of moderate-size sensor networks. Similarly, the choice of established networking standards like IP (Internet protocol) provides a wide choice of deployment and management tools (e.g., DHCP for addressing, DNS for naming, SNMP or HTTP/SOAP/REST for management), thus removing the need to reinvent everything.”

Augusta Systems:
“Clearly, costs of a sensor network will increase as the number of nodes increase, but if the network design is intelligent and open, these increases should be limited to the hardware costs of the nodes and not enhancements to the network architecture. However, in a poorly designed system or in a system where the network integration and data convergence issues are overlooked, the cost and complexity can increase exponentially.”

Digi Intl.:
“It depends on whether or not the underlying network structure changes at all. If it is a point-to-multipoint network and the new nodes use the same central host, then the cost increase is limited to the cost of buying and installing the new hardware, and the increased complexity does not come into play. If, however, the increased node count requires a change to a mesh network, then there may be different types of nodes in the network that have to be thought about and managed. Timing and data delays may change as well.”

Dust Networks:
“Larger networks do not necessarily mean more costly networks. With adequate forethought into bandwidth usage and traffic patterns, a self-organizing network can be scaled quite large with simply a linear relation in cost (e.g. 10 nodes = $x 40 nodes = $4x) and no increase in complexity. Complexity and cost rise rapidly when point-to-point networks scale since network connections must be carefully engineered to withstand all eventualities.”

Eka Systems:
“The size of the network can dramatically impact the overall cost, complexity, and the very viability of large-scale sensor networks. The networking and routing technology is inherently complex by an order of magnitude (compared to) small, simple networks. There are many flavors of networking technology that work for a few thousand nodes, but they totally breakdown when tens–of thousands of nodes are involved in a single network. Dense mesh-networking technology to handle hundreds–of thousands to millions of nodes in one network should be architected from ground up. Scalability should be the starting point and not an after thought. Clever routing algorithms, automatic configuration, quick route convergence, knowledge of the cheapest and best link at all times, across-the-network reachability with the smallest routing table while conserving processing, and memory resources are the critical characteristics required to create, monitor, manage, operate, and maintain million-node sensor networks. Metering application deployments are the true test for validating such large-scale scalability with very high reliability. To be sure, there is not a single network anywhere on the planet with a hundred thousand nodes yet. The vast majority of deployed networks to date have a few hundred to a few tens of thousands nodes.”

Ember:
“The complexity of a sensor network increases with network size. However, the robustness of the standards—in particular the ZigBee PRO Feature Set approved in 2007—is effectively insulating users from the underlying complexity. The per-node cost of such ZigBee systems does not change with the size of the network however, a larger network may require additional tools to manage ongoing operations.”

GE Sensing:
“For mesh networks in general, the cost and complexity decrease on a per-node basis as the number of nodes increases. For simple star systems, smaller is both better and cheaper.”

GreenPeak Technologies:
“The cost of a wireless sensor network consists of the fixed cost for the basic infrastructure and the number of sensor nodes needed to cover the monitored area. The cost increases linearly with the number of sensor nodes installed. More careful planning of implementing parallel wireless sensor networks would be required for a network containing more than 1,000 sensor nodes.”

Jennic:
“The complexity of a sensor network does not necessarily need to increase with an increase in the number of nodes. Most intelligent sensor network deployments utilize clusters of wireless sensor networks (say up to 100 nodes per cluster) that are connected through a wired backbone. The wireless clusters are inherently more manageable and can be used extremely effectively to distribute network loading and RF (radio frequency) activity. The wired backbone gives you guaranteed message delivery through the hop from one cluster to the next.”

Mesh Systems:
“We believe that cost and complexity scale relatively linearly with the number of nodes. In small sensor network applications (10–30 nodes per gateway), the cost of the gateway becomes significant. In larger applications, the gateway cost is less of an issue, but the overall complexity of managing reliable communications with a large number of end points can be significantly higher.”

Millennial Net:
“The key is not the number of nodes in a mesh network but the number of ‘“hops”’ and the ability of the mesh protocol to be able to dynamically reconfigure itself quickly on a continual basis.”

Pedigree Technologies:
“This is highly dependent on the sensor networking technology employed. The sensor-network nodes can vary in cost depending on the application requirements placed on the sensor network. Generally, we have found that those sensor network technologies that work hard to get information back to a gateway/base station in one to two hops are very competitive in terms of cost and managed complexity. Those technologies that require multiple “short” hops to capture the data generally require more hardware nodes depending on the placement of the gateways or access points.”

Sensinode:
“By making proper use of available wired infrastructure in conjunction with wireless access points, the number of nodes can be increased without a large increase in complexity by breaking down the problem into smaller pieces. This is a large part of the value of IP standardization, which will allow organizations to more easily leverage existing infrastructure and IT (information technology) department experience with IP networks. Large mesh networks (few access points) have problems with power consumption, coordinator complexity mobility, and performance that can be minimized with an IP-centric approach.”

5. Management
What is required to manage a sensor network on a daily basis?

Arch Rock:
“Users need a network infrastructure that inherently provides management capability (e.g. naming, addressing, monitoring, alerts, thresholds, etc.) and provides it in realtime—not based on a preset schedule as some approaches require. Sensor networks should not be islands of information, but an integral part of the larger enterprise information infrastructure the two must be able to ‘speak’ a common language. That is why existing IT (information technology) standards are so important. If, for example, the sensor network speaks IP (Internet protocol) and can be integrated into the IP network, it automatically ‘inherits’ the vast body of IP tools for interoperability, management, access control, etc.”

Augusta Systems:
“Once a sensor network is properly installed and deployed in a manner that ensures intelligent convergence and integration of data, very little effort is required to manage the sensor network. Problems occur, however, when sensor networks are not integrated with the enterprise IT network. For example, remote sensor networks may require manual data gathering in some instances where network connectivity is unavailable, and thus, there would be an increase in network management costs.”

Digi Intl.:
“Little, if any, daily management is needed once the wireless networks are setup. There are tools that can allow for remote monitoring, but normally these types of tools are used on an as-needed basis and they are not required for day-to-day operation.”

Dust Networks:
“For a sensor network to be economically viable, there should be no daily requirements to manage it. The purpose of wireless is to simplify the tasks of monitoring and control, if it adds its own overhead this goal is compromised.”

Ember:
“The management of the sensor network on a daily basis is highly application dependent and is tied in with the management requirements of the overall system of which the network is a part. For example, in a home automation or security system, customers may need to monitor battery life of battery-powered devices (much like checking batteries in a smoke detector or entertainment remote control). In a commercial building automation system, the network is monitored as part of the overall HVAC (heating, ventilating, and air-conditioning) or lighting system, within which the mechanical systems likely need more attention and maintenance than the communications.”

Eka Systems:
“Sensor networks that require daily management cannot scale and are not viable. Only self-healing, auto-configurable, self-managed sensor networks can be deployed to scale. Sensor networking technology should focus on just being there, highly available, adapting to the operating environment at all times, creating the best links and maintaining the reach of a dynamically changing operating environment. Technologies requiring intense network management on a daily basis are not appropriate for sensor networking. However, end-to-end service management and using data visualization tools is essential to deliver and maintain end-user services.”

GreenPeak Technologies:
“A seamless integration with existing IT is required to manage the wireless sensor network and its measured data. All sensor networks have built-in diagnostics and status feedback, which can be used to by management software to generate reports and trigger alarms that can minimize the required management tasks.”

Mesh Systems:
“Ideally, the ongoing management of a sensor network is performed automatically by gateway and server software, and reporting and notification of problems is done only on an exception basis. Communication retries, heartbeats, self-healing, and server-side intelligence are all important in minimizing ‘'noise’ from sensor-network systems.”

Pedigree Technologies:
“In our experiences with various deployments, most commercial-grade sensor networks usually don’t require daily management. The importance lies, of course, in the availability of good system software that will provide administrative mechanisms to ensure network health. The standardization of sensor network technologies should enable more widespread system software tools to address maintenance issues. It would be of great interest to the market-at-large to see SNMP-like (simple network management protocol) platforms make their way rapidly and very deeply into the sensor-networking realm. Likewise, it would be very good for the senso-r network standards bodies to welcome ubiquitous SNMP-like functionality into their families.”

Sensinode:
“The management of a sensor network can and should be made to look like that of any other IT infrastructure management. Standard tools then can be used to manage access points, and wireless node management is typically built into the application-specific service. Wireless nodes should be designed for full autoconfiguration, allowing remote management through the infrastructure.”

6. Troubleshooting
Who is typically responsible for fixing problems with a sensor network after deployment?

Accsence:
“Shared responsibility between ourselves and the customer.”

Augusta Systems:
“Generally, sensor networks tend to be managed and maintained by network or system administrators and engineers. Ideally, troubleshooting should be no different than that for other IT (information technology) systems.”

Coronis Systems:
“Network maintenance depends on the service agreement, typically between an integrator and the end customer. It is advisable that organizations using a sensor network have a trained staff, while specific applications or niche solutions may require the ongoing expertise of an integrator. No system is maintenance free.”

Digi Intl.:
“The company that sold and installed the systems is responsible if issues arise after deployment. Most OEMs (original-equipment manufacturers) and other parts manufacturers will provide a supporting role in resolving problems.”

Dust Networks:
“A self-organizing sensor network requires minimal troubleshooting which can be performed by onsite technicians with no specialized wireless knowledge. If specialized wireless consultants are required to maintain a sensor network, it will not serve the business needs of the end user.”

Eka Systems:
“Troubleshooting starts at the end-user NOC (network operations center) and could end up at third-party services providers towards final resolution. The customer/end-user owners are typically responsible for all Level-1 troubleshooting across all the three main components: the backoffice business systems, the backhaul, and the deployed sensor network. Application, data-manipulation, data integrity and other issues related to data in the back-office is the responsibility of backoffice software vendors, system integrators or end-users themselves. Backhaul connectivity issues, when isolated to data delivery, are the responsibility of the contracted service provider. Network issues beyond application and data delivery are resolved by the network technology provider under support contracts.”

Ember:
“The communications network is typically seen as a feature / component of the system of which it is part. Hence, the service provider (installer/dealer for a high-end home-automation system, the building maintenance staff or vendor for building automation system, or the consumer themselves for home-area networks of low-end systems) would have the responsibility.”

GE Sensing:
“Depending on the nature of the relationship, either the installer or the supplier will be responsible. But, as a general rule, the installer will work on the wireless connectivity, while the supplier will work on the software and sensor functionality.”

GreenPeak Technologies:
The company which installed the sensor network is most often responsible for technical support. The level of support will typically be defined in a SLA (service-level agreement or maintenance contract.

Jennic:
“Generally speaking, this would be the responsibility of the computer network personnel. Therefore, systems that offer standard and well-understood mechanisms for fault diagnosis are clearly desirable, ping and trace route being obvious examples for identifying routing problems in IP-based networks such as 6LoWPAN (IPv6 over low-power wireless personal-area networks).”

Millennial Net:
“The integrator or installer typically takes this responsibility. However, often the customer also assumes this role. The skill set requirement is that of typical sensor maintenance personnel. In some cases, IT personnel may include WSN within their domain. With the ability to monitor sensor network performance remotely via the Internet, expert troubleshooting support has become greatly simplified.”

Pedigree Technologies:
“With the proper training most end users can troubleshoot problems in coordination with the sensor network integrator or OEM. Given the fact that there are end-to-end wireless sensor network solutions being rolled out in markets such as home automation, it is fair to say that the complexity of setup, maintaining, and troubleshooting is light in many applications.”

7. Interface
How is sensor data usually presented to the end user?

Arch Rock:
“Sensor data should be presented in any format the end user wants, whether via standard web services APIs (application programming interfaces), running standard SQL (structured query language) queries against a database, or downloadable to a spreadsheet via CSV (comma-separated value) format. Presenting data in common, standards-based ways gives the user an almost unlimited choice of how to view the data at any given time.”

Augusta Systems:
“Presentation is usually dependent upon the needs of the end user and can include stand-alone applications or integration with other enterprise systems. Generally, we find that most enterprises require sensor-network data to be integrated into existing enterprise systems.”

Digi Intl.:
“Ultimately most end users want to view and monitor data on a PC (personal computer). That means that even if the network is comprised of analog sensors, somewhere along the line the data must be converted to a digital format so that it can be received by a PC via Ethernet, RS-232, USB, Wi-Fi, or whatever backhaul method the customer wants.”

Eka Systems:
“In applications such as industrial controls, sensor data is presented in a dashboard with a graphical interface specific to the application. In metering applications, sensor data may not be directly presented at all, but rather digested, validated, and summarized by backoffice software systems before being stored in databases. The data is usually stored in a universal data format that can be easily and quickly integrated with backoffice business systems or transparently manipulated by external off-the-shelf reporting and data visualization tools.”

Ember:
“The end user is generally looking to achieve some function that is dependent on the application. This could be anything from turning lights on/off, checking to see if the alarm system is armed, or controlling the temperature in a room/building. Hence the user interface is as varied as the systems and could include a Web-based interface on a PC, a thermostat, or in-home wall-mounted display, a message received on a PDA (personal digital assistant), or a two-way entertainment remote control.”

GreenPeak Technologies:
“The sensor data gathered in the network is transferred via gateways, through the existing IT (information technology) infrastructure, into database structures. This information is made available to the user in application specific client applications or through Web pages in a browser. These applications often contain algorithms and artificial intelligence—also called sensor reasoning algorithms—to manage the vast amount of data captured in the sensor network, and present it in a form directly useable by, (for example) a building management operator. Sensor information can be transformed into reports and alarm events, on which human interaction is needed.”

Jennic:
“Sensor data can be represented in many different ways depending upon what the application wishes to see. Ultimately the data from a sensor node will be a binary value that represents the value of a particular node variable. The way in which this value is represented will largely depend upon the required interface into the business system.”

Mesh Systems:
“Over the Internet to browser clients, scheduled reports emailed in printable form, scheduled reports in spreadsheet form, or the sensor information database exported to other business applications. For notifications, we have found that most users prefer email and/or SMS (short message service).”

Millennial Net:
“Typically an embedded control application, software application or portable device.”

Pedigree Technologies:
“In many applications, the sensor data is married to specific applications so the presentation of raw data is limited. The extension of sensor networks to Web 2.0 applications make for interactive, view-anywhere usability that is gaining traction for many of the new asset management applications. Otherwise, classical HMI (human-machine interface) and graphs, tables, and charts become the standard interface to look at data or information.”

8. Integration
How can companies integrate sensor data with other business systems?

Arch Rock:
“Standard Web services, standard IP (Internet protocol) networking, and open interfaces in general are the de facto tools used for communication by today’s business systems, and there’s no reason that sensor data should be integrated in any other way. These tools have proven to be the most scalable and efficient means of achieving integration between highly diverse sources of data, from financial data to sensor data.”

Augusta Systems:
“Integration of sensor data into enterprise systems is significant. Standalone applications tied to one vendor or device can create silos of information that may impede the distribution of data throughout the network. The use of common protocols and file formats to both receive and send data can break down these troublesome silos.”

Eka Systems:
“Sensor data in universal data format from sensor networks can be quickly integrated with existing relational database engines like Mysql, Oracle, DB2, and many others. Existing software development tools are then used to create or modify business systems and end-user applications to use the integrated sensor data. Adopting backoffice Web-service standards across application verticals enables easier integration and enhances the value of the data.”

Ember:
“Typically various sensor data is aggregated by a gateway of some type that is linked to some sort of backend IT (information technology) system. A simple example might be a home-awareness system consisting of motion, light, door/window, and temperature sensors that communicate to a home gateway connected to a user’s broadband network. The gateway may then communicate to a service providers’ backend system that can be accessed by the consumer via the Web or mobile device to ascertain the status of his/her home systems. A similar setup could be envisioned for an industrial tank-monitoring system that might use GPRS (general packet radio service) as a means of communicating aggregated sensor data to a central monitoring station.”

GE Sensing
“OPC and other standard interfaces provide the best means of integrating sensor data with other business systems.”

GreenPeak Technologies:
“In its most common form, sensor data is stored in databases the integration with existing IT infrastructure can be done by accessing that data—via standard IP-based services and infrastructure—to feed intelligent software which can extract relevant and useable information which can be used in existing business systems.”

Jennic:
“Protocols such as the IP or UDP (user datagram protocol) can present sensor information in a widely understood and globally adopted format that can be easily taken and integrated into existing IP-based business systems. These protocols are already deployed on a global scale, and use within wireless sensor networking simply reaches that step further, for a fairly seamless implementation.”

Mesh Systems:
“Some companies have implemented Web-service/XML mechanisms for accessing and integrating sensor data into other business applications. Others simply want sensor data exported in standard database formats.”

MeshNetics:
“In order to integrate data from WSN (wireless sensor network) to an enterprise system, a company will need a gateway solution that would consist of hardware and software/middleware parts. Hardware platform will normally feature USB and Ethernet interfaces to connect WSN and IP networks, although Wi-Fi and GPRS gateways also exist. Software part becomes increasingly more important because of the emergence of the multiple IEEE 802.15.4-based standards (ZigBee, 6LoWPAN, SP100, WirelessHART), and the existence of a relatively large number of proprietary protocols.”

Pedigree Technologies:
“Typically, organizations want to look for companies that provide flexible middleware systems with standard integration points that are uniquely designed for sensor network or intelligent device-network applications. Then it is purely up to the business systems ability to utilize and present the information to the end user in a meaningful way.”

Sensinode:
“Currently the ideal way to integrate sensor data into other systems is using a SOA (service oriented-architecture). The sensor network must be able to send data to business systems using SOAP (simple object access protocol) messaging, as that is the basic building block of most modern business systems.”

9. Duration
How should adopters approach a temporary installation compared to a permanent one?

Accsence:
“Not much difference with a wireless system.”

Arch Rock:
“Wireless technology should be able to handle both temporary and permanent installations. Of particular concern is ensuring that the solution supports autonomous power and communications that can be configured quickly and with minimal effort. In addition, IP-based (Internet protocol) WSNs (wireless sensor networks) are able to natively traverse multiple network links regardless of the medium and thus are particularly well suited to mobile, autonomous, ad hoc deployments.”

Digi Intl.:
“Every system should have some preliminary testing in a lab environment to at least know if the various pieces of the system will communicate together before they are deployed in the field. With a temporary installation, normally the network can be pieced together without as much concern for long-term maintenance issues. With a permanent installation, design so that maintenance is minimal is a key factor.”

Eka Systems:
“Early adopters should use pilots as a way to learn the nuances of the network to create the data set of most value. Pilots are a great way to focus on the business must haves while separating the nice to haves. Data gathered during this process validates the going-in assumptions and captures the true network and application requirements for a permanent installation.”

Ember:
“This is completely application dependent. Wireless sensor networks make fast, temporary installations that much easier and simpler to implement.”

GE Sensing:
“Temporary installations are much easier in wireless networks since the customer can move around nodes to create a viable network and gauge the performance.”

GreenPeak Technologies:
“When using wireless sensor networks, there is no real difference between deployments of temporary or permanent networks. This contrasts with wired solutions, for which a temporary installation is cumbersome. The wireless mesh-network technology is self forming and self healing: nodes will automatically find alternative communication routes when signal quality decreases or when the infrastructure changes, again saving time and money.”

Jennic:
“When the tools and techniques for a temporary installation are those that you will use for the final installation, then it makes perfect sense to use the same approach for both deployments. The practice of carrying out a site survey, followed by installation, followed by performance tuning will be required for both. Albeit you will want to spend more time on the site survey for the temporary installation to get initial node placement correct, and the process of performance testing will be more thorough for the final deployment, but ultimately the process should be the same for both.”

Millennial Net:
“Other than different battery-life selections, there is essentially no difference in approach.”

Sensinode:
”Permanent installations can make better use of, or demand new, infrastructure components such as access points and local servers. A temporary installation should make use of available laptops or PDAs (personal digital assistants), combined with the use of technology, for example GSM (global system for mobile communications), to make it location independent.”

10. ROI
What are the best ways to measure return on investment for a sensor network?

Accsence:
“Through three factors: labor savings, product loss savings, regulatory compliance savings.”

Arch Rock:
“Overall investment for sensor networking is fairly straightforward and determined by the system costs (e.g. sensor nodes, routers and servers, and corresponding software), installation and configuration, and finally, ongoing operation, maintenance, and change management on the system. The return is where things get interesting. The benefits accrued from the new visibility created by the sensor network can be enormous (e.g. preventive data for asset maintenance, failure avoidance for process monitoring, accident avoidance for personnel safety, asset tracking for loss or damage avoidance, energy awareness for consumption reduction, etc.). Not as easy to quantify, but sometimes ‘priceless.’

Augusta Systems:
“As sensing needs increase and as deployments become more complex, there are a number of factors that impact ROI (return on investment): personnel costs for both initial deployment and management network infrastructure costs (such as bandwidth challenges resulting from the sensor network) and data convergence and integration costs. The ultimate measure of ROI, though, is whether the enterprise is receiving operational benefits relative to these costs. Generally, the ROI will be high if the deployment and management is simple, the burden upon the communications infrastructure is low, and the convergence and integration of data with other system and network data is easy. However, even in complex deployments, ROI can be significant due to the savings from having better operational insights on business conditions and minimizing manual data collection or estimation efforts.”

Dust Networks:
“ROI varies widely by application and industry. In many industries reducing worker risk, increasing productivity, and cost reduction are typical ROI measurements for wireless.”

Eka Systems:
“The true value of a sensor network is measured in the productivity and enhancement of business applications and operations driven by realtime data delivered reliably from a sensor network. The ROI is a measure of gains across multiple levels in any application environment. The network provides the ability to create and retrieve new data and intelligence that did not exist before. Command and control improves business responsiveness that can quickly translate into new revenue generating opportunities. Integration of sensor data with existing business applications enhances the value of the services provided while reducing cost and improving efficiencies. Networking reduces day-to-day operations and field support costs of geographically dispersed assets. Initial deployments will be largely confined to cost reduction initiatives. As deployments mature, realtime and historical data will become more valuable and drive new service offerings and future revenues.”

Ember:
“Again, this is more a function of the overall system in which a sensor network is part. Wireless sensor network technology such as ZigBee can dramatically impact the overall system and installation cost by removing the cost of wired installation.”

GE Sensing:
“On wireless networks, the total hardware cost is normally slightly higher than that of a wired system. The user needs to take the total costs into account (Hardware + cost of installation + cost of maintenance) when looking at his return on investment.”

GreenPeak Technologies:
“For general sensor networks, ROI can be measured in increased efficiency, higher security, energy savings, etc., depending on the application. Also the economies in labor cost for automatic sensor reading versus manual measurements can be enormous.
For wireless sensor networks, the extra savings can easily be established after calculating the wired and the wireless deployment savings, where cable cost is avoided and installation cost of the sensor nodes is minimized.”

Jennic:
“There is no fixed science for calculating ROI. However, when you consider the ease of which you can replace a node, with which you can introduce a redundant or additional repeater node, the time saved from limited or no wiring, the ability to diagnose network routing issues or packet loss symptoms through standard ping or traceroute commands, then all at once the metrics considered for a ROI calculation need to include not only basic system costs, but the savings through reduce operational/diagnosis/maintenance activities.”

Mesh Systems:
“If an adopter truly understands the costs (hardware, installation, and operating) and can quantify the benefits (performing a function better, faster, or cheaper), then the measurement is actually easy and no different than any other return on investment calculation. We have found it is far better to focus on directly measurable benefits (such as reduced cost) than non-quantifiable ‘pie-in-the-sky’ benefits.”

Millennial Net:
“ROI needs to be measured by the value the data at the application level to improved business performance. The sensor network is merely a tool to enable the data transfer, and as such, there is no different approach to measuring business value than with any alternative investment of a remote monitoring and control application.”

Sensicast Systems:
“Sensor networks are in operation in a wide range of facilities and provision a wide range of applications (especially in measuring physical and process conditions). ROI is both application-specific and also determined by the strategic and operational imperatives of the operating companies. Users of sensor networks say they are tracking and have seen rapid and significant ROI in the form of: savings of direct costs like electricity and materials process-throughput improvement increased quality lowered total cost of ownership of the sensor network deployment of a wireless sensor network eliminates the higher cost and
headaches of installing wires and cables ability to run new applications that were not possible without wireless.

Sensinode:
“In application-specific cases, ROI should definitely be measured in process improvement where sensor networks have great potential, for example in asset management. In more horizontal-technology cases, ROI should be measured by comparing alternative technology solutions for the same problem.”

PART 2: TECHNOLOGY QUESTIONS

1. Resources
What level of technical skill is required to set up and maintain a sensor network?

Digi Intl.:
“In most instances installers do not have to be as technically savvy as IT (information technology) personnel, but they do need to be somewhat computer literate and have an understanding of the manufacturers test and configuration software.”

Dust Networks:
“For a sensor network to be economically viable, it must be simple to set up and maintain using commonly available tools (varies by industry) and require no specialized wireless skills. In industrial applications, any competent instrument technician should be able to install wireless sensor networks. In commercial building applications, the same is true but for HVAC (heating, ventilating, and air-conditioning) technicians.”

Eka Systems:
“Networking expertise is essential to making the most of large commercial and industrial sensor networks. Technical skills acquired in the computer networking industry are invaluable in setting up and maintaining a sensor network. Radio expertise is equally important in demanding application environments with interfering and often competing wireless technologies. In many instances, a single such resource with rich industry experience is more than sufficient to maintain a very large network of thousands of sensors.”

Ember:
“The level of technical skill required depends on the end application and the complexity/scale of the network. In the case of wireless sensor networks for home automation, there are DIY (do-it-yourself) solutions available in the market today. However, in applications with a larger number of nodes such as in advanced metering , high-end home automation or building automation, professional installation is typically required for set up. It should be noted that for AMI (advanced metering infrastructure) systems, the communications from the meter to the utility is utility-owned and installed, while the home-area network of in-home devices would typically be installed directly by the consumer.”

Jennic:
“The level of technical skill required depends upon the requirements of the application and also the tools and application interfaces that can be provided by the technology vendor. Developers will typically be provided with an application interface that is highly abstracted enabling them to focus on the value add within the application rather than getting involved in the intricacies of the networking stack. Equally, they may choose to get involved with the intricacies of the stack for finer control by interfacing at a lower level. With this in mind, the process of setting up a network can be an activity that takes as little as a few hours out of the box, or it can be a many-month development exercise.”

Millennial Net:
“The skill set requirement is that of typical sensor/instrument maintenance personnel. In some cases IT personnel may include WSN (wireless sensor networking) within their domain. Again, remote Internet-based network performance monitoring greatly simplifies remote support.”

Pedigree:
“In some cases, the skills to setup and maintain the network can be done by laymen, especially if the network is small and the application is not critical in nature. The larger the deployment and the more critical the application, the employment of wireless integrators combined with existing network administrators for maintenance should do the trick in most cases.”

Sensinode:
With IP (Internet protocol) standardization, the resources already available in a typical IT department will be sufficient to deploy a large commercial sensor network. For small, e.g. home-based, sensor networks a technical-savy user should be able to set up and make use of nodes using standard IP protocols.

2. Setup
How is a sensor network installed and activated?

Arch Rock:
“A sensor network is installed and activated by deploying a number of sensor nodes along with a mechanism for connecting them back to an aggregation or collection point. Mesh networking allows each node to determine the best path back to an egress point (e.g. router), either directly to the point itself (if it is in close proximity) or by multihopping via other nodes. The most flexible and reliable networks are those that make it possible for each node to act as a router for another node’s traffic.”

Dust Networks:
“The installation of a sensor network varies by vendor and technology. For a self-organizing network, an installer will first install a gateway and then install wireless devices as they would with a wired product. The network will then establish device-to-device and device-to-gateway connections automatically based on realtime conditions.”

Eka Systems:
“The components of a sensor network are: the sensor nodes, gateways, backhaul links and backoffice business systems. To deploy a viable sensor network that can grow at will and over time into thousands of nodes, it must be self-activated and install itself when powered up and form the network without any human intervention. Gateways act as data collectors and route data from a large group of sensors to the backoffice using a single backhaul wide-area link. Gateways are factory configured with hardware and software for a specific backhaul service provider in addition to operating like any sensor node on the network side.”

Ember:
“This is highly application dependent. A home security system can be simply unpackaged by the consumer and set up with a few simple steps. A large commercial lighting control system requires professional installation and commissioning of the various devices (i.e. assigning specific controllers to specifics lights). Other systems are as simple or complex as the total system requires.”

GE Sensing:
“The gateway/base/hub is turned on first and then all the sensor nodes are turned on and installed in their respective locations. As soon as the nodes get powered up they start communicating to the base and form into a network. The network goes on continuously at the network layer to keep the network healthy.”

GreenPeak Technologies:
“A wireless sensor network is self-forming and the nodes are auto-configuring. Once a node is powered, it will automatically link to the wireless network. This ‘plug and play’ approach enables every technician to execute a 100% correct installation without the need to understand technical aspects of setting up a wireless network.”

Mesh Systems:
“We typically use a push-button or magnetically-operated reed switch to initiate installation and registration of a sensor end point to its gateway. The same switch can be used to signal end point removal or replacement. Feedback to the installer is provided by an LED indicator on the node. Gateway activation can be done in advance of the physical installation or through Web-based forms.”

Millennial Net:
“Typically all of the devices are installed and connected to power if appropriate and then activated. The integrated gateway device is then installed and automatically finds all devices. This information is then used as the basis for configuring an end application (such as for example energy management). Installers are uninvolved with any networking or system configuration..”

Pedigree Technologies:
“Many sensor network applications of the wireless nature can be done in a fairly ad-hoc nature with a good centralized software platform to commission and manage the sensor network devices.”

3. Logistics
What are the range limits and physical limitations of a sensor network?

Digi Intl.:
“For a wireless system, the output power and frequencies used are regulated by government agencies. In line-of-sight applications, even low-power radios can transmit a long distance, but obstructing buildings, trees, or antennas mounted close to the ground can all significantly reduce the range of a system.”

Dust Networks:
“The range limits and physical limitations vary by vendor and technology. Typical sensor network applications are relatively densely spaced (2-20 meter distance). Regulatory agencies limit the power output of wireless devices in the license-free ISM (industrial scientific medical) bands which in turn limits range.”

Eka Systems:
“Sensor data is typically very small, just an order of few bytes and the communication is never continuous or chatty. These characteristics define a sensor node with low-data rates integrated with highly efficient low-cost radios and microcontrollers. The radios operate at very low power to reduce component and hardware cost. Sensor nodes are hence, range limited. Based on the application environment, some nodes can communicate up to a mile. However, most sensors are expected to operate within a few hundred meters. This distance limitation imposed by hardware is overcome by mesh networking and routing software to extend the range using other sensor nodes as routers and extending the network reach to thousands of nodes and large distances. Memory, processing power, and battery are other physical characteristics that dictate the operating parameters of sensor nodes. In battery operated environments, efficient processing and low footprint code is critical to maintaining nonstop operation for years. Small and efficient code keeps a lid on expensive memory chips.”

GE Sensing:
“Depending on the cost that a user can afford, sensor networks range from 500 feet to 50 miles.”

GreenPeak Technologies:
“Single-hop communication range varies between 10 meters and 75 meters, depending on the building infrastructure and the building materials used. Outdoor single-hop limitations are somewhere between 100 and 1000 meters, also depending on the obstacles.
However, by implementing a wireless mesh or multihop network, where messages hop from one sensor to another between source and destination, one can do away with these range limitations completely, enabling the network to span a virtually unlimited area.”

Jennic:
“Range limits come about through the characteristics of the environment and the selection of the sensor node. All environments are different and they all have their unique range of challenges, such as steel reinforced concrete floors providing characteristics analogous to a Faraday Cage, moving people and machinery that may create momentary RF path blocks or interference and interference from other wireless devices.
When looking at sensor nodes, they will typically achieve ranges from 30 to 50 meter indoors and in line-of-sight conditions, they can easily exceed a few kilometers of range. Equally there are nodes with maximum power options, different types of antenna to choose from, and antenna diversity mechanisms, all of which help to improve the characteristics of the sensor node.”

Mesh Systems”
“It depends, of course, on the particular physical medium chosen for the application and available power. We have found 868/915 MHz solutions to be generally better than 2.4 GHz for long-range applications. For battery-powered end points, operating at 868 or 915 MHz, we typically plan on 500 meters (open air) and 100 meters (in building) as the maximum hop distance. Antenna design is critical.”

MeshNetics:
“There are multiple factors that affect the range, such as environment conditions, interference, and obstacles, that may affect the range, making it difficult for a manufacturer to claim any particular range in their technical specifications. Furthermore, when a ZigBee module is integrated into a specific device with a particular antenna, the performance may change. A typical range of a 2.4GHz ZigBee module is 10-30 meters indoors and about 300-400 meters outdoors line of site. Use of a power amplifier increases range three or four times.”

Millennial Net:
“Advanced radio platforms with 100mW of power can adequately operate with a range of 500 feet in typical environments.”

Pedigree Technologies:
“The limits are usually expressed in “trade-offs” you can get for really long range, but you will pay for it in more frequent replacements if the nodes are battery powered. Localized power on a portion of the network is sometimes required. Short-range wireless applications can improve battery life but sometimes require more hardware deployments, thus increasing cost. The best is to find a provider that has optimized range with limited hardware necessity within the context of your application environment.”

Sensinode:
“High-power 802.15.4 2.4 GHz radios can achieve several kilomeers of range line of sight and up to 100 meters indoors. By using limited amounts of multihop and covering smaller areas with access points linked to a backbone network, the problem is broken down and can scale to huge numbers of nodes. Bandwidth (e.g. 250 kbps in 802.15.4) is a limitation for nodes in the same place, which can be improved with multiple overlaying networks and future 802.15.4 generations.”

4. Power
What should an adopter understand about power requirements and battery life?

Accsence:
“Battery life is a complex equation and depends at least on: signal quality, sensors driven by the battery, number of nodes, position in the mesh, and data transmission frequency (measurements/time).”

Digi Intl.:
“For batteries to last multiple months or years, energy saving and sleep modes need to be used. The amount of data that needs to be transmitted and how often those transmissions have to occur along with the sleep current, will determine the power consumption. The average power consumption metric will help the adopter size the appropriate battery. It is also important to understand that some repeaters or mesh routers may need to use a wall outlet or some other permanent power source as not all technologies allow for sleeping routers.”

Dust Networks:
“Battery life should be viewed in terms of the application requirements and typical equipment maintenance cycles. Proven sensor-network technologies exist today that deliver over a decade of battery life for both end nodes and router nodes.”

Eka Systems:
“Optimal battery lifecycle management is critical to the overall longevity of individual sensor nodes and hence the entire network. Application requirements directly impact power and battery life. Early adopters should be fully aware that power-aware networking software and application software is essential to deploying a large-scale sensor network. In a completely battery-operated mesh sensor network, clever wake-up and sleep algorithms keep the network humming for years. This duty cycle should be designed with specific application data requirements to minimize power drain.”

Ember:
“The important thing that adopters have to understand in wireless sensor-network applications is that the sensor/transceiver uses most power when it is transmitting or receiving data. In order to extend the battery life, the sensors spend most of the time in ‘sleeping’ mode. For example, ZigBee end devices may implement various “sleep” modes in order to allow years of life with low-cost batteries.”

GE Sensing:
“All the wireless products in design these days are very low-power devices and have significantly high battery life. At a one-minute sample rate, the sensors normally last more than a year on most short-range networks.”

GreenPeak Technologies:
“A wireless sensor network typically contains two types of devices: mains-powered devices, with unlimited power supply, often called ‘full-functional devices’ or router nodes, and so-called end devices or reduced functionality devices, running of a limited power supply such as a battery.
Battery lifetime of such nodes varies between a couple of months and about 10 years, depending mainly on the type of battery and the communication rate at which the device needs to communicate, as well as temperature and other parameters influencing the natural self-discharge and capacity of a battery. Highest quality standard batteries such as AA or coin cell lithium batteries have a maximum lifetime of about 10 years, which puts a hard limitation to battery lifetime.
When battery lifetime does not exceed the lifetime of the appliance itself, battery-powered devices create a maintenance problem due to the need for frequent battery replacements. This problem can be tackled by implementing ultra-low-power wireless sensor networks, which can be powered by energy harvesting (batteryless applications powered by a solar panel, electromagnetic, or vibration harvester, etc.) or equipped with coin cell batteries that last longer than the sensor’s estimated lifetime.
In some cases, powering devices from mains is not a viable option. When there is a need for a multihop mesh networking running off batteries, low power routing technology.”

Jennic:
“For battery life the most important thing to understand is the duty cycle of the application. Most sensor networks will typically have many nodes that spend most of the lifetime sleeping and then periodically wake from sleep to take a measurement and broadcast the result to a data collector. In this situation, the dominant factor for battery life tends to be the sleep current. Sure, the environment can affect the system, specifically if the sensor has to perform multiple retries to send a message through, but typically for a sleep current of less than 5µA with a sensor pole of something more than 30 seconds, then an application ought to see multiyear battery life from a reasonably sized battery of say 1000mAhr.”

Mesh Systems:
“Battery life is always approximate and depends on sensor sampling rate, polling interval, event reporting requirements, distance from nearby nodes, and network topology. We have found that an investment in high-quality batteries specified for use over a wide temperature range (such as lithium thionyl chloride) is generally worth it. Battery capacity measurement techniques should be implemented in all battery-operated devices and battery condition periodically reported.”

MeshNetics:
“Wireless sensor networks are generally expected to run autonomously for a few years without having to replace batteries. The rule of thumb here is the longer the devices remain in sleep mode, the longer the battery life will be. Using the chipset with lowest power consumption in sleep mode will help achieve longer battery life.”

Millennial Net:
“There is a direct relationship between battery consumption and radio power. Typically for a given power level the only way to reduce battery consumption is by decreasing the “duty cycle” or sampling rates, which typically means decreasing the responsiveness and performance of the network.”

Pedigree Technologies:
Flexible network layouts are an ideal. Pure mesh networks require nodes to route traffic route from other nodes. This requirement requires a great deal of intelligence in the network, which subsequently greatly reduces battery-life applications. An example of a nice network layout is a tree configuration. The branches should be routing-capable nodes that can be plugged into a line power source. This frees the sensor nodes (the “leaves” of the tree) from routing responsibilities and allows them to achieve much longer battery life.

Sensicast Systems:
“Network performance criteria based on power-related issues depends on the frequency of data measurement and the design of the network protocol and topology. Typically smart sensors (preintegrated probes with battery-powered radios) can continually conduct measurements in one-minute intervals for two to three years between battery changes. Systems that include AC-powered mesh routers as part of the solution will positively impact battery life.”

Sensinode:
“A common misconception is that radio listening is free, whereas it usually costs as much as transmitting. For this reason, low-power wireless nodes should sleep most of the time and off-load listening to powered access points. Nodes performing multihop forwarding are heavy energy consumers and need careful planning. With proper design, wireless-node power consumption can be kept to a long-term average of under 100 uA.”

5. Security
How can an adopter make sure its sensor-network data is secure?

Accsence:
“All data should be over-the-air encrypted.”

Augusta Systems:
“Appropriate encryption is essential. Sensor data security tends to come in four forms: structured within kits, as components within integration toolsets, as elements of the communications system components, and as add-ons within sensor-network software code.”

Digi Intl.:
“There are a number of different levels of security, but if data security is a concern then it is good practice to use products that support a tested, standard encryption algorithm. For many wireless protocols 128-bit AES encryption is used.”

Eka Systems:
“Network and data security is achieved in a sensor network with multiple levels of defenses. Authentication, encryption, and access control collectively protect sensor networks. Authentication is the process of identifying and validating the communication initiator or the source sensor node. Authentication protects compromised nodes from corrupting the network. Encryption is the process of converting data into a coded form to prevent it from being read and understood by unauthorized entities. AES-128 standard is the most widely used encryption standard and requires enormous resources to tamper. Access control is the process of limiting the communications to certain information, services or functionality based on the identity and profile of the authenticated sensor node.”

GreenPeak Technologies:
“The wireless sensor-network standards like IEEE 805.15.4 and ZigBee include security features by default. Encryption in the hardware blocks and security services and algorithms are deployed at no additional cost and require no additional power consumption.”

Jennic:
“By using inline encryption techniques, adopters can securely encrypt data. AES-128 encryption is provided with IEEE 802.15.4 standards-based chips or solutions, and in some instances a dedicated on-chip hardware AES engine is offered, providing an almost zero-overhead inline processing capability. The AES will typically be used for all message communication, with a user specific encryption key of 128-bits that makes the information interchange highly secure.”

Millennial Net:
“In proprietary mesh networks, typically the complex algorithms provide a strong measure of embedded security within the mesh network. In many applications this is considered an effective barrier to network intrusion. Additionally, most advanced radio platforms also support hardware encryption for additional security, with the tradeoff of a reduced network efficiency. Standard AES security models are considerably more important for mesh network standards due to the open and published nature of the protocol, and thus are always an inherent component of any standard.”

6. Reliability
What steps should an adopter take to ensure reliability of the network?

Accsence:
“Make sure the signal quality is good. This is the key factor.”

Arch Rock:
“RF (radio frequency) communication is difficult and governed by the laws of physics, so reliable networking is never a given. Reliability should be visible to the end user and easily monitored. Users can improve reliability by choosing an architecture that allows for multiple edge devices (e.g. routers) this introduces a level of redundancy, thereby making the network more robust. Furthermore, any vendor should be able to show users in realtime how a network is performing.”

Coronis Systems:
“Product pilot testing is the best way to iron out reliability problems. Wireless technologies today are generally able to communicate as advertised, so reliability is frequently a question of making sure you are using the right technology for your application. This, plus automatic alerts sent from the field when network anomalies are detected, are good ways to ensure network operation in the long run.”

Dust Networks:
“Given that the strength and quality of any one RF link is unpredictable, adopters need to make sure they have a network that can route around temporary blockages or interference. Without the ability to route on different paths, use different frequencies and retry on failed communications, network reliability cannot be achieved. Adopters should be sure they are using wireless networking technologies which take advantage of spatial, frequency and time diversity. This will ensure extremely high reliability in the face of harsh RF environments.”

Echelon:
“Any wireless sensor network that is expected to perform with a reasonable degree of reliability or that must have consistent response times must include a wired backbone for all the repeating points, at a bare minimum.”

Eka Systems:
“Sensor-network reliability is an inherent function of the radio, networking, and routing technology embedded in the sensor nodes. Frequency-hopping radios perform much better than single-frequency radios that operate using collision-avoidance schemes. Whether it is a ten-node network or a ten-thousand-node network, the reliability expectations remain the same. A smaller, reliable network is highly vulnerable to breakdown as the network expands if not engineered to scale from ground up. Choosing a scaleable routing technology is a critical component of ensuring network and data-delivery reliability.”

Ember:
“The topology/architecture used can greatly impact reliability. A mesh-network topology is inherently self-organizing and self-healing. It offers multiple links among nodes, so messages automatically travel along any available link, greatly enhancing the reliability of the overall network. In a mesh network, extra nodes can be quickly deployed to serve as repeaters to fill in any holes in the network.”

GE Sensing:
“For a mesh network, it is critical to ensure as many neighbors to a node as possible for better reliability. It should be mandatory to at least have two neighbors to ensure reliability.”

GreenPeak Technologies:
“Reliability is related to the availability or absence of a communication path between two wireless devices. The most critical enemy of reliability is wireless interference coming from other users of the same wireless frequency band.
Wireless sensor networks have built in a whole range of measures to overcome interference such as over-the-air collision avoidance, so called CSMA/CA (carrier sense multiple access/collision avoidance), or ‘listen before talk’ algorithms, where sensor nodes check the medium for other communication before it starts sending itself. Another way to battle interference is the addition of redundancy in communication paths, by using the meshing principle to make networks self-healing. Moreover, at network start-up, a mesh network will automatically search for the ‘most quiet’ channel, and can adapt its operating frequency during operation (so-called frequency agility) to avoid new sources of interference.
Adopters need to make sure that all of these measures are implemented in the technology used. In most cases, these interference measures are automatically executed and do not require any manual intervention from a skilled technician.”

Jennic:
“There are simple measures that can be taken during initial network deployment to aim to provide the greatest opportunity for a network to be reliable over time. Measures such as signal strength or packet error rate between two nodes can be observed and these will give an immediate indication as to the likely reliability of a particular point-to-point link.
Suffice to say, however, that over time the environment is likely to change and therefore a network that supports a degree of self-healing capability can provide additional robustness to help with long-term network reliability.
Frequency agility also provides a solution that helps with a changing environment, but a degree of caution should be used here. Continuous tracking of channels to find the clearest is at the expense of battery life. Equally, nodes within a network will likely suffer in different ways—what may be a good channel for one node, may not be a good channel for another due to localized interference.”

Mesh Systems:
“To start, adopters need to recognize that RF communication is not 100% reliable and that great care must be taken to ensure that the network can withstand temporary loss of communications between sensor end points, gateways, WANs (wide-area networks), and centralized servers. Installation processes, whether performed automatically or manually, should be designed with sufficient communication link quality margin to allow for degradation of the communications paths over changing conditions. Communications-link statistics between communicating devices, such as RSSI (received signal strength indication), error rates, retry counts, etc., should be maintained, analyzed, and reported. Network topologies such as mesh that support self-healing and alternate communications path formation are also important in ensuring network reliability.”

Millennial Net:
“Reliability is primarily determined by the protocol itself. In a real-world environment, self-forming, self-healing, low-latency, and highly scalable protocols, which are able to ensure 100% packet delivery integrity (no dropped packets, inherently deliver the highest level of reliability provided there is: (a) path redundancy and (b) general bandwidth availability within the network for the sensor data. Thus, adopters should ensure deployments are designed with all devices having at least two neighboring devices within range and that the overall sampling rate of the devices does not overwhelm the network bandwidth limitations.”

7. Infrastructure
How can a company leverage its existing network infrastructure when deploying a sensor network?

Augusta Systems:
“If intelligently designed, a sensor network should maximize the existing network infrastructure—communications standards, application architectures, network hardware and other components. The sensor network should exist as the edge element within an existing IT infrastructure, instead of as a parallel infrastructure or stand-alone application.”

Digi Intl.:
“A lot of system parts can be used together. It is fairly easy to take a Modbus network that is wired using RS-485 and expand the system to difficult locations using radios for those outlying nodes. In-house testing prior to deployment will allow you know if you can use them in the field if the need arises.”

Echelon:
“The best choice would be to hang clusters of wireless sensors off of I/O (input/output) blocks that are part of an existing building-control network and leverage the built infrastructure, including the IP backbone.”

Eka Systems:
“Sensor networks should be envisioned as the first mile geographically dispersed wireless microLANs (local-area networks) interconnected with gateways. When economically feasible, existing end nodes can be retrofitted with wireless sensor-networking hardware and software mesh technology to fully leverage deployed assets. Sensor networks are sources of old and new data which can be designed to transparently map to existing backoffice systems and applications, while providing a rich data set for future applications. Sensor networks should be engineered to ring in new ways of getting data to existing infrastructure. Companies can re-engineer existing backhaul and backoffice infrastructure to connect and communicate with sensor networks.”

Ember:
“This is application dependent. The application profiles defined within the ZigBee standard allow integration with existing control networks within the different application areas. For example, the emerging Commercial Building Automation profile is compatible with the BACnet protocols used in many building control systems.”

GE Sensing
“If there is an existing Wi-Fi network in place, it is possible to tap into this with wireless sensors. However, our experience is that this is very infrequent.”

Jennic:
“The obvious answer for how to leverage existing network infrastructure is to use gateways that bridge the various network technologies and protocols. Gateways can bridge different hardware technologies such as wired to wireless, and bridge network protocols such as ZigBee to IP. The hardware conversion generally speaking is the easy bit, the network protocol conversion can be somewhat complex and so begs an obvious question: Is there a common protocol throughout the existing network that could be adopted within the sensor network, and if so, what is it? Once this is known then the decision should be to find wireless technology, ideally offering the same or something very close.
Many of the existing network infrastructures today are based on IP for information technology, for business, for field bus, for building management systems. The specification for 6LoWPAN (IPv6 over low-power wireless personal-area networks), offers wireless sensor networking based on IP for data communication. This can make the exercise of systems integration very straightforward and seamless. A wireless sensor network can become just a simple extension of the existing IP network.”

MeshNetics:
“Various gateway solutions make that possible. If a wireless sensor network is set up in a remote location, a GPRS (general packet radio service) gateway can link it with the main network. Inside one building, an Ethernet bridge or RS232 or USB interfaces can be used. If, for instance, a wireless sensor network is set up on each floor of a multistoried building, an Ethernet backbone can link the disparate networks together. Software is an important component of such gateway solution.”

Pedigree Technologies:
“The use of protocol translators such as gateways and access points are the best ways to use and conform to existing network infrastructure.”

Sensinode:
“Ethernet infrastructure, including the use of PoE (power over Ethernet), is an excellent place to start. With a good Ethernet network, it is easy to add Ethernet-to-802.15.4 routers as a base for the deployment of sensor networks. Existing servers can and should be used also for running sensor network services and for the management of the sensor network.”

8. Hardware
What criteria should adopters use for selecting a hardware platform?

Accsence:
“Complex question. We suggest getting something that works, turnkey, without regards to standards, and hold the vendor accountable from end to end.”

Digi Intl.:
“Adopters should ensure that the hardware they are using has the right certifications and is approved for use in the countries or environments where they are deploying a system. In addition to that, expected ranges, supporting products the manufacturer offers, and warranties should all be considered so adopters can meet their customers’ expectations.”

Eka Systems:
“Apart from cost, size, and application-aware functionality, zero-touch-configuration and zero-touch-scalability are essential attributes required to deploy a successful sensor network. zero-touch-configuration creates a reliable sensor network. This capability is the process of deploying a factory-shipped sensor node in the field, turning it on, and walking away with the full knowledge that the network will get formed automatically, and the data will find its way through the network and end up in the backoffice systems. Zero-touch-scalability enables ‘at-will’ network growth transparently. This capability enables adding and replacing nodes without impairing or bringing down any section of the network, thus avoiding expensive down-time and potential loss of revenue generating data.”

Ember:
“Adopters should take into consideration factors such as standards, product reliability/performance, and cost. They should also consider the success the platform of choice has had with other adopters in different applications and environments.”

GE Sensing:
“Distance between the nodes, power requirements, and cost affordability are some of the factors to be considered.”

GreenPeak Technologies:
1. Standards based which brings reliability, interoperability, and second sourcing flexibility
2. Energy efficiency, enabling maintenance-free sensor networks (no frequent battery replacement or even batteryless)
3. Cost effectiveness: an integrated solution that consists of an integrated microprocessor, energy management, and application-specific sensor interfaces”

Jennic:
“Without question adopters obviously need to ensure that the hardware platform supports the features required by the application and almost always at a price point that is acceptable or better than the expectation. However, buying into wireless sensor networking is much more than a simple hardware selection process.
Questions that need to be asked are:
- What software support is offered?
- What network stacks and applications interfaces are provided for the hardware platform?
- What about evaluation kits and development tools?
- What level of technical expertise do I require to work with the technology? (Do I need RF (radio frequency) design experience? Are there module solutions to get me to market quickly?)
- Are reference designs, both hardware and software, readily available?
- What about technical support, forums or direct, for field-applications engineers?
Adopters should be fully aware that they are buying into a total technology solution and not just a hardware platform.”

Mesh Systems:
1. Adequacy for the intended function
2. Cost
3. Financial and technical strength of the provider(s) and their ability and willingness to provide support.
4. Field proven in related or similar applications
5. To a lesser extent, whether or not the platform is based on standards”

MeshNetics:
“When selecting a ZigBee radio, one must look at its link budget. The combination of Rx sensitivity and Tx power is called ‘link budget,’ and is related to the range of operation. The difference in link budget—just nine dBm—nearly triples its range. Power-consumption characteristics of hardware are also extremely important for autonomously operating sensor networks. There are other factors that determine the choice of the hardware platform, i.e. size of the chipset, computing ability, analog-to-digital conversion and peripherals, development environment and compliance to regulations (RoHS, FCC, CE, etc.).”

Millennial Net:
“The need for extremely low power and efficient yet complex networks, such as mesh networks, means the hardware design is closely coupled with the protocol being used. When selecting a hardware platform, there are several critical factors pertaining to power consumption including sleep and power-up modes and isolation of hardware functions, such as watchdogs and synchronization. Rapid advances in radio, microprocessor, protocols, and power-harvesting technologies are driving improvements in cost, performance, and size. These are happening independently by multiple suppliers and driven by markets beyond WSN (wireless sensor networking). Therefore, at the current level of maturity of the WSN market it is advantageous to avoid a hardware design that has closely coupled these technology elements into a rigid platform (e.g. system on a chip).”

Sensinode:
“In selecting a hardware platform, the cost of the chip(s), power consumption, performance, standards, open HW interfaces, multicompiler support, flash programming tools, vendor availability, and the availability of commercial protocol solutions are the most important factors.”

9. Intelligence
When should a sensor network have distributed processing capability?

Accsence:
“When communication is unreliable.”

Augusta Systems:
“Ideally, sensor networks should always have distributed processing capabilities since these capabilities enable optimal data convergence and integration with other systems and the enterprise IT (information technology) network. The increasing use of sensor networks can create communications bandwidth constraints. Distributed processing can help to overcome these constraints.”

Digi Intl.:
“It is needed when capabilities are expected to expand. If the data format is not critical, then the sensor defaults may not need to change and additional intelligence may not be required. Expanded capabilities, such as advanced battery management, data parsing or filtering, or data-format changes, generally will require some processing capability.”

Dust Networks:
“Distributing processing capability can help reduce overall network traffic and increase battery life. If an edge node can analyze raw data and only send on event, then battery life can be increased.”

Eka Systems:
“Devices that require peer-to-peer application conversation (example: light and a switch) are good candidates for distributed processing at the application level in sensor networks. Star-type data conversations always end up at gateways or collection points and require no application processing at intermediate nodes. However, at the network- and routing-protocols level distributed intelligence is already built into most mesh sensor networks.”

Ember:
“Any ZigBee wireless sensor network platform inherently contains distributed network processing capability. At the next level, the need for distributed sensor data processing and analysis is a function of the sensor system architecture and independent of the network operation.”

GE Sensing:
“For any networks that need more than 50 feet, it is desirable to have distributed processing capability.”

GreenPeak Technologies:
“By default, wireless sensor networks are built on distributed processing capabilities. Each sensor has its own microprocessor managing its own reliability in the network (end device) or looking after neighbor nodes (FFD, fully functional device or router).
Indeed the price will determine the amount of processor capacity of each sensor node.”

MeshNetics:
“Any large wireless sensor network must be resilient enough without increase in latency and risk of the packets’ loss. This can be achieved by implementing mesh routing. On-the-node computations can significantly reduce the unnecessary traffic within the network, improving battery life at the same time. For example, sensors monitor parameters such as temperature, humidity, illumination, etc. However, an operator often needs to maintain a combination of parameters, e.g. ‘room comfort level,’ that would include certain combination of temperature, humidity and illumination values. The raw sensor data can be therefore preprocessed on the node before being transmitted.”

Millennial Net:
“There have been extensive studies on the optimal distribution of intelligence within a distributed processor environment demonstrating that intelligence should be applied at the lowest level at which the ability to make a decision exists. This strongly suggests that distributed processing be applied to sensor networks especially when there is an element of control required.”

Pedigree Technologies:
“Distributed intelligence capabilities usually are a function of the critical nature of the application, the network limitations of the LAN (local-area network) or WAN (wide-area network), battery-driven intensive requirements, and even the capabilities of the backend. Realtime applications, especially those that include multiple network devices and implement control features can require distributed processing. If your WAN is secure, power is plentiful, the applications are not critical and the data coming in is not overwhelming, nothing beats cost-effective powerful servers can handle a lot of work.
It is also important to note that some network devices should maintain their specific identities and job sets. In the hands of an overzealous engineer/developer what once was an intelligent gateway has now become a “server,” which leads to an increased chance for failure and unnecessary network and software lifecycle complexities.”

Sensicast Systems:
“Sensor networks should ALWAYS have distributed processing. Some networks are more centrally controlled than others. Self-healing, self-configuring sensor networks are inherently more reliable. A ‘distributed nervous system’ is also better for scalability and redundancy.”

10 (a). Standards
What should adopters understand about the standards associated with sensor networking?

Accsence:
“That there are no good standards that mean much to an end user.”

Dust Networks:
“Viable sensor networking standards are here today. In the home automation space Zigbee is gaining traction, and in the industrial markets WirelessHART has been ratified and products are expected to be on the market this year.”

Echelon:
“Sensor networking standards are still in their infancy. Interoperability has yet to be fully defined or formed as yet. The development of interoperability standards is essential for any networking protocol. Without them, wireless sensor networks that deliver an enduring competitive environment to end users—one where end users have a rich choice of vendors, integrators, and management software at inception and over the life of the system—won’t exist. Instead, end users will be putting in wireless sensor networks that are built by a single vendor, service by a single provider, and extending only by the original suppliers.”

Eka Systems:
“Technology adoption usually happens years before standards mature. Technologies in various stages of ongoing standards development exists in deployed networks. Deployments with the capability for adopting to new standards with over-the-air software downloads offer the most flexibility and robustness to stretch the dollars in a rapidly evolving sensor-networking industry.”

Ember:
Standards-based sensor networking solutions enable adopters to have a choice of vendors, as well as interoperability between products from different vendors. Adopters also need to understand true standards, which entail a choice of vendors at the networking platform level, as opposed to proprietary networking platforms used in multiple product offerings.

GE Sensing:
“The wireless standards at this time are still in their infancy. SP100 is currently being promoted by most of the prominent companies.”

GreenPeak Technologies:
“Wireless communication prospers best within the space of industry standards. Standards offer OEMs (original-equipment manufacturers) the freedom to purchase from a larger pool of suppliers. Additionally, standards allow devices from different vendors to interoperate, a feature which is paramount in applications ranging from building automation to industrial automation.
Several standards are currently available, but there is no ‘'one size fits all’ standard available to suit the totally different requirements most wireless networks have. Several standards currently coexist and share the PHY (physical) and MAC (medium access control) specifications of the IEEE 802.15.4 standard.”

Jennic:
“Where available, a standards-based solution should always be chosen as it is more likely to stand the test of time, and equally, you can be reasonably guaranteed that there will be multiple vendors offering equivalent products, therefore avoiding the dangers of being locked into a specific vendors technology. Standards play a key role in driving the adoption of wireless sensor network technology. The standard for wireless sensor networks is the IEEE802.15.4, and many implementations already exist that support this standard.”

Mesh Systems:
“Adopters need to look beyond the ‘hype’ associated with certain sensor-networking standards and carefully assess the true benefits that adoption of a standards-based solution brings. What is the extent of the standard? Does it cover only physical/MAC layers, or (is it) designed to provide true application-level interoperability? Is interoperability required or even desired? Is the standards-based solution more cost-effective than a proprietary approach? Is the standards-based solution available from multiple technology/component vendors?”

MeshNetics:
“IEEE802.15.4 is a major standard associated with wireless sensor networks. It serves as a base for such higher-level standards as ZigBee, SP-100, and WirelessHART. Its physical (andMAC provide the foundation, although do not determine the networking. A higher-level standard ensures compatibility and interoperability of devices for various applications. Currently, ZigBee is the only open standard that addresses a broad variety of wireless applications. Each particular application, such as lighting control, automated meter reading, or HVAC (heating, ventilating, and air-conditioning) control, demands a distinct set of functionality. Therefore, ZigBee developed separate profiles for different groups of applications that define the network devices and protocol of communication between them.”

Millennial Net:
“Standards are taking longer to establish themselves in proving their commercial viability in part because users and developers need to ensure they adequately meet the specific needs of the application for which they are being deployed. WirelessHART is an example of where a standard has been designed to meet a very specific process monitoring need and when deployed within its target market should prove itself to gain more rapid acceptance than other prior standards.”

Pedigree Technologies:
“Standards are very important to the process because of the tremendous potential of the applications both classical and new that these technologies will address. These standards allow for smoother integration with other networks, systems, and devices. Standards will also be needed to achieve the many new pervasive computing applications that the pundits speak of. On the flip side, adopters should understand that standards take time to deliver and implement, but they are worth it.”

Sensinode:
“IEEE 802.15.4 is the most important place to start, it provides a globally interoperable and largely supported radio. Although ZigBee provides vertical solutions for some specific applications, a lack of profiles and the closed nature of the SIG (special interest group) means that adopters may not find it to be optimal in many applications. The most open and universal solution is currently the 6LoWPAN (IPv6 over Low-power wireless personal-area networks) IPv6 standard which provides an alternative to proprietary protocols over 802.15.4.”

Sentilla:
“The most important thing to understand is that the value of a standard is directly proportional to the number of persons who use that standard. Standards bodies may define criteria and certify a ‘standard,’ but if it is never adopted by the community, then the standard itself is irrelevant. Adopters should whenever possible embrace standards already used by millions rather than trying to promote new standards that have yet to demonstrate traction among users.”

10 (b). More Standards
Which proprietary protocols should adopters avoid?

Accsence:
“None.”

Arch Rock:
“Users should always pressure their vendors to provide standards-based solutions. Of particular note are proprietary radio technologies that are available from a single vendor, as these make up a high proportion of the customer’s hardware investment that would be wasted if the vendor could not sustain a viable long-term business or compete cost effectively with standards-based alternatives.”

Dust Networks:
“If interoperability and multivendor support is required, all proprietary protocols should be avoided.”

Eka Systems:
“Proprietary protocols that are not designed to scale from small networks to tens of thousands of nodes are the ones to avoid. Such protocols lack the theoretical foundations to build, converge, and maintain very, very large networks. This issue is not so apparent in smaller networks, as there are many proven proprietary solutions that are economically feasible. However, for very large networks the considerations as stated are essential to deploy a viable solution.”

Ember:
“Ember is a strong proponent of a standards-based approach for the long-term success of the technology as well as consumer ‘future-proofing.’ As such, proprietary protocols are rarely a good investment from a long-term perspective.”

GE Sensing:
“Most of the currently available protocols are proprietary, and the wireless standards are still evolving. Ensure that the proprietary protocols you choose are reliable, support frequency hopping, and network security. Lastly, make sure that the provider has a plan to migrate to one of the wireless standards and has a means to upgrade your system.”

GreenPeak Technologies:
“In addition to the IEEE 802.15.4 standard, a number of technology suppliers have chosen to build a proprietary transceiver. The main motivation seems to be a reduction of the complexity and thus a potential lower cost point. However, it remains to be seen if a proprietary solution will ever reach sufficient volumes to actually reach that theoretically lower cost point. Additionally, reducing the complexity automatically goes hand in hand with sacrificing performance and thus limiting the applicability. A proprietary protocol could be a suitable solution in an early-adopter stage of market development, but in the future, all proprietary protocols should be avoided.”

Jennic:
“Suffice to say that proprietary protocols are generally an inevitable consequence of evolving standards. They shouldn’t be faulted, but should be considered with an edge of caution, and actually in some instances, they may bring added value that is unattainable from the totally standards-based solution. The use of proprietary application protocol or network protocol may mean that a system is essentially closed to the outside world—this may be beneficial for security reasons.”

Millennial Net:
“It would be a mistake to avoid considering proprietary protocols as these can provide significant performance benefits. However, it is important that the protocol is self administering and responsive in a dynamic environment. A track record of proven deployments that are aligned with the adopters requirements is key to avoid unexpected deployment (and lifetime) costs.”

Sensicast Systems:
“Two standards should be avoided: (1) ZigBee and (2) TDMA (time division multiple access).”

Sensinode:
“Most non-802.15.4 radios should be avoided except in fully closed applications with special requirements. Single-vendor proprietary protocol solutions are something to also be wary about. Even protocols like Z-Wave with fair support will have limited lifetime as they will be replaced with real standards.”

Related Topics: Connected Home

Related Stories: > New GPS Module Enables Tracking; > Minimizing the Cost of Operation; > Global Sustainable Road-Mobility Management; > M2M Reduces Liability from Distracted Driving; > A Community for Developers

Comments

or Create Account to comment on this, or other articles: