BeyeNETWORK - The Global Vision for BI and Beyond

The Evolution of RFID System Architectures and the Development of Pervasive Computing

Passive Tags and Active Tags, Centralization and Decentralization

Published: February 26, 2008

RFID tags are on a path to becoming the basic building blocks and the volume driver of pervasive computing.

Although people currently think in terms of “RF” (radio frequency) tags that “ID” (identify) and track products and other assets through supply chains, RFID tags will eventually be thought of as very small computers that happen to have a built-in wireless networking capability. In other words, it is only a matter of time until virtually every item with a RFID tag can become a programmable/intelligent node on a private extranet, a private intranet, or the public Internet. What this means is that RFID tags are on a path to becoming the basic building blocks and the volume driver of pervasive computing.
 
The evolution of RFID system architectures will, in many respects, be like previous IT evolutions that saw competition between centralized and decentralized designs. With RFID, as tags become recognized as programmable computers with finite but ever-increasing amounts of processor and memory capacity, IT architects will engage in competition to develop system solutions with the appropriate trade-offs between “tag-oriented” and “network-oriented” system designs.
 
Both system design camps will argue that Moore's Law is driving computing MIPS closer to free while the network-oriented advocates will remind us that Moore's Law is also having a similar impact on conventional bandwidth and web-based servers.
 
In addition to total life cycle cost (one-time and recurring costs), various architectural issues including RF range, power consumption, functionality, flexibility, reliability, redundancy, security, system management, scalability and ease of use will drive the competing system designs.
 
Innovative and aggressive tag-oriented system architects will advocate squeezing ever-greater capabilities onto each RFID tag. This will be the course for most active tag suppliers and potentially some passive tag suppliers. The tag-oriented architecture proponents, especially the active tag suppliers, will pursue applications including relatively long-range access control, high-value asset management and various other applications requiring true real-time locating systems (RTLS) capabilities. Initially, more so than the network-oriented proponents, these suppliers will also be the likely providers of sensor-equipped tags.
 
Sensors will become an increasingly important function. Sensors will be used to monitor and report changes in the environment such as temperature, humidity, shock and vibration, and various aspects of security such as physical or other tampering. Sensors on RFID tags will detect and report chemical and nuclear particles. The applications for RFID tags with and without sensors will be nearly endless.
 
Network-oriented system architects will shift as much functionality as possible to the infrastructure of web-based networks and servers and will have a strong (huge) initial focus on passive tags. Over time, as costs fall, the network-oriented advocates will add sensors.
 
Network-oriented architects will say, "Why add any functionality on the tag that could possibly be put on a web-based server? Just get the cost per tag down as low as it can go and get the volume of tags as high as possible as quickly as possible." The tag-oriented architects will respond by saying, "But unless you put this or that functionality on the tag, the application won't work – no matter how much you want to use web-based servers."
 
In both the tag-oriented approach and the network-oriented approach, extraordinary advances in flexible miniaturization will drive impressive on-tag functionality including standards-driven, tag manufacturer-driven, and user-driven data fields, programmable processes and other features.
 
Over time, as RFID tags become recognized as what they are – miniature programmable computer nodes on a network - the distinctions between the tag-driven and network-driven architectures will recede. At this point, RFID architecture will resemble other IT architectures that have matured to offer granular increments of processor and memory capacity with options. However, where previously network interfaces on computers were considered the front end, on RFID tags they will become considered the back end and sensors will become the front end.
 
In the early days, tag providers will offer very specialized tags. As the market matures, the winning providers will offer a family of tags that will support a high degree of functional portability among the tags. The tags, the readers (reader-writers), the ability to manage the RFID edge environment and the ability to integrate the edge environment with IT systems everywhere will give rise to platforms for pervasive computing.
 
As a frame of reference, some analysts (focused on conventional computers, mostly PCs) have said that the one billionth computer was shipped in 2002 and that the two billionth computer would ship by the end of 2007. While forecasts are certainly subject to error, projections and reports have indicated that by the end of this year (2008), a total of nearly 8 billion RFID tags will have been shipped. No doubt it is easier to make projections than sales; and clearly most of the tags shipped to date have likely been used in access control, smart card and supply chain management applications. However, it is important to remember that increasingly RFID tags will not be just devices to identify and/or track items, but rather they are on the path to become programmable and networkable computers. The applications for RFID will be limited only by increasingly easier levels of cost-justification and people's imaginations. It is worth repeating: virtually every item with a unique serial number (or name) will have the potential to become an intelligent wireless node on a private intranet, a private extranet or the public Internet.

How will such a pervasive computing architecture evolve? It is hard to predict in detail – although it is likely to evolve in a manner that is driven by vertical industries and the automation of horizontal workflows – but it seems safe to say that pervasive computing will consist of some combination of centralization and decentralization.
 
Centralization and decentralization are basic concepts. Centralization invests more in resources at the center of the architecture (think of an expensive mainframe at the center with relatively inexpensive dumb terminals at the edge of the network). Decentralization invests less at the interior and moves more intelligence – perhaps at greater cost – to the edge of the architecture (think of PCs decentralizing computing after years of centralized mainframe-based computing).

The economics that drive centralization or decentralization are dependent on many assumptions. It is difficult to provide a formula that says which is better, but what the market will often prefer is something that provides the lowest total cost of ownership (TCO). TCO for an information system must include everything needed to make the solution work (hardware, software, networking/bandwidth, security, installation, training, support, taxes, overcoming a financial hurdle rate, etc.). In other words, TCO must consider all the one-time and recurring costs for the life of the project.

However, it is important to note that the overall economic equation that drives centralization versus decentralization is not solving just for the lowest TCO, but rather it often attempts to solve for the lowest TCO that provides the intended functionality. Sometimes the cost of the adopted solution is higher than the cost of one or more alternative solutions, but the higher cost solution is adopted because in addition to a higher cost, it also brings greater benefits. In such a scenario, a larger investment results in a larger return on investment (ROI).

At the end of the day, there are only two ways to increase profit: decrease costs and/or increase revenues. Any architecture that is driven by business will need to show one or both of these financial benefits.

It might be interesting, as a case study, to look at solving various problems with passive RFID tags versus active RFID tags. Passive tags (analogous to a centralized architecture in which relatively little (unit) cost and intelligence is ascribed to passive tags on the edge) rely on considerable centralized server resources to compile lots of data into actionable information; most of the computing is done centrally. Active tags (analogous to a decentralized architecture in which relatively more (unit) cost and intelligence is ascribed to active tags on the edge) rely on battery-powered tags with more powerful on-board semiconductors and perhaps sensors to process and record relatively more data – effectively pushing more intelligence to the edge of the decentralized architecture.

So which is better?

It all depends on what is being addressed – what is the challenge or opportunity and what is the time frame in which the challenge or opportunity must be addressed? The best business case usually wins – so it is often helpful to model the TCO and ROI for alternative approaches. In the end, the cost of a RFID tag (whether passive or active) is only one factor in a multi-variable equation. History has shown that both centralized and decentralized architectures can prevail for extended periods of time in the marketplace.

This article is published with permission from RFID Recruiters, LLC.