Computing Fabrics: The BIGGER Picture
Part 1: Computing Fabrics Extend the Cell Phone Paradigm
October 26, 1998

Our presentation of Computing Fabrics has so far depicted scalable and flexible computing power that can be more effectively and efficiently utilized than today's "rigid" technology. Applications in the data center, such as ever larger data warehouses and OLTP systems will be early adopters of Computing Fabrics, driven by their need for more processing power that is also more affordable and more manageable. However, new applications, qualitatively different than any today, will depend on Computing Fabric's unique properties, not just their efficiencies.

These future applications supporting greater mobility and collaborative work will include sophisticated next generation user interfaces featuring elaborate 3D visualization, robust speech recognition, language and knowledge processing, intelligent agents, and computer-based modeling of the user - anticipating the users' needs and preparing appropriate responses in advance. Progress to next generation user interface technologies will mandate another leap in processing power within or very near the user interface, just like what enabled the evolution from the command line to the GUI.

Computing Fabrics can deliver this local power by extending a paradigm subscribed to by millions every day when they place or receive a call on their cell phone. Let's take a look at the cell phone paradigm and two closely related paradigms from computing. Then we'll return to Computing Fabrics and how they could build on these paradigms to support next generation applications.

Cell Phones

In today's fast-paced world most of us have personal experience with cell phone technology - we freely move about, on foot or in a vehicle, and are apparently connected with the entire telephone network. Should we drive out into the country or the desert our connection becomes weakened, sometimes even severed. Other times our connection strength decreases and then gradually resumes. The reason? Cell phones possess only limited transmit and receive power because their designers properly assumed that they would only exchange signals within a local cell, defined by the placement of antennas, often in a grid. We are wirelessly connected by a cell phone only over a short range with the local cell - which cell is local depends on where we are, which changes over time as we move. The local cell is then connected into the telephone switching network by other means, be they wires, fiber optics, or microwave. As we move about we connect into a local cell which, being part of the larger network, provides us global connectivity. So when we depart a heavily populated area (which has regularly laid out cells) and enter the boonies we are farther from a cell antenna and therefore our connection grows weaker. Although this simplified picture glosses over numerous complexities it hits on the essentials - connect locally (wherever "local" happens to be), communicate globally

Network Computing redeploys this paradigm within the context of computing.

Network Computing

A highlight of Network Computing (unfortunately overshadowed by wimpy network computers) is access to our personal workspace - our desktop - from anywhere on the network, be it across corporate LAN, WAN, or the Internet. Once we have access to our desktop we can then access any of the resources that are available to us from our desktop. Need to check your calendar from a tradeshow, find a contract while visiting a client, or access data from the corporate database late at night from a hotel room? All of these are potentially doable provided that the desktop applications that provide access to these services are small, downloadable, operable across a wide variety of platforms, and importantly, that essential configuration information cannot only be accessed but is automatically invoked. Though this is a demanding list it conceptually paints the picture of a very desirable property - transparency of location. Just as the cell phone enables us to talk to Tokyo whether we are creeping along Madison Ave. in downtown Manhattan or four-wheeling through a dried-up riverbed in Arizona, the Network Computing concept provides us with our personal computing space wherever we are.

Ubiquitous Computing takes this concept and makes access even easier and more direct.

Ubiquitous Computing

Ubiquitous Computing, most commonly associated with research at Xerox PARC, depends on processors, displays, and input devices being omnipresent - everywhere office workers are likely to be. Each office worker is tracked throughout the office complex by electronic, magnetic, or ultrasonic sensors. This information is then utilized to direct each workers' applications and data to the appropriate devices and displays that are located where the worker is located, wherever that is as the worker moves throughout the complex over the course of the day. The devices can be large displays, electronic whiteboards, PDAs, or even standard workstations. A worker's session is therefore almost always within arm's reach. I reference applications, data, and sessions, instead of a desktop, because in this highly mobile environment a "desktop" becomes somewhat meaningless - no longer chained to a desk the entire office complex becomes your workspace. (We'll expand on this specific concept in the next two columns.)

Ubiquitous Computing improves on Network Computing in at least two ways. First, the user need never dialup or login to access their computing space - it's just always there, wherever they are in the office. Second, in distinction to the cell phone paradigm (which enables the user to connect with the world from anywhere), Ubiquitous Computing brings each users' digital "world" to wherever they happen to be, ultimately expanding that digital world throughout the entire office and taking the first steps at coupling the digital world with the office.

Issues with the prevailing Paradigms

While the cell phone paradigm has eased life for many of us (and upped our phone bill!) the same cannot yet be said for Network Computing and Ubiquitous Computing. While cellular merely requires that we utilize a small interface device - a cell phone with charged batteries or the equivalent - these other two, Network and Ubiquitous Computing, require the user's possession of computational resource that are up to the task of whatever application(s) the user wishes to run locally. The marketplace has already rejected network computers as lobotomized last generation PCs. A modern client needs local horsepower, lots of it, even when the network services accessed are well supported on huge back-end servers. As for Ubiquitous Computing, it is still a lab experiment, requiring large investments in workstations, displays, and input devices that need to be located simply everywhere.

How can next generation applications be supported - mobile - collaborative - verbal - 3D spatial - intelligent applications - without generally over-providing capacity to each user in the form of a portable supercomputer the size of a cell phone? The answer are VPSCs - Virtual Personal Supercomputers that can self-assemble out of the local processing nodes and interconnects of a Computing Fabric to meet the needs of these next generation applications.

We'll explore these VPSCs next week.

Erick Von Schweber