Networks and Distributed Systems
Research group of Prof. Peter Bernard Ladkin, Ph.D.
Future University Computing Resources
Peter Ladkin
Research Report RVS-RR-06-05Predicting the future is hard. But predicting the past is even harder (as a good dictionary shows). The computing environment changes fast. Not only may current capabilities be augmented many-fold in a few short months, but fundamentally new uses arise every few years, as do styles of use. It is comforting to realise that no major use has fallen out of favor, even if styles of use change frequently. But in light of fundamentally new uses, the past is no reliable guide to the future. How should a University plan for future computing needs?
Distinguishing Needs from Styles
Universities must support infrastructure for teaching and research. Computing infrastructure is dependent on the uses to which computers are put, and a mistaken choice can tie up millions of Deutsche Marks or other real money and become ineffective in two or three years. The good news is that the needs of computing users are foreseeable--they exist now and will remain stable. However, in most people's minds, needs are tightly bound up with local details which are part of the style of use, and which become passé as new solutions emerge. It is thus fundamentally important to distinguish needs from the mechanism provided to fulfil them, which I shall call the style of use. It is harder to predict what style will catch on to meet a need.For example, take the WWW. For years, researchers have been `surfing' the Internet to search for and retrieve documents which interest them. We have been awaiting tools to aid us for many years, convinced that they would arrive. So they did -- in the guise of a very well-designed concept of how the net would appear when the tools were used. Web surfing caught on because it was a simple, clear concept, directly supported by portable, simple tools, to fill a pressing need: to read and retrieve public documents held on the Internet. The need was there, and already recognised. The style was provided by the WWW-tools from CERN. They might have failed, had they not been so apt, or if the need had not been so pressing. Given that it was, some solution was bound to be found.
A distinction between needs and styles is not hard and fast. Since one can describe situations at many levels of abstraction, one can imagine roughly a hierarchy of concepts in which styles at one level are needs at a lower level--for example, communication of documents is a need, WWW access is a style; given that WWW is a need, the style of communication at a site may be either radio or wired transmission of information; wired transmission may be considered a need for a University, and FDDI+backbone or ISDN/ATM may be a choice of styles.
Infrastructure Needs
Infrastructure needs should be considered at a level which is consonant with planning. There are two aspects: time and materiel. Four-year plans require four-year abstractions. And noone can afford to rebuild buildings too often to accomodate the latest trend in wiring.Wiring offices for terminals for interactive computing became important in the late 60's/early 70's and is unlikely to go away in the near future. Interactive terminals have turned into desktop computers, but the wiring needs are similar. The major break came with Ethernet. Another, still underway, with FDDI. Yet another may come with ISDN/ATM, in which phone lines may prove adequate for much, although existing FDDI/Ethernet cabling will remain technically superior. There is a commercial question as to whether FDDI will extend from LANs to MANs. Wires may be supplanted by infra-red and radio communication, but current projections of cost suggest that wires will remain the medium of choice for a decade or more: radio-communication companies face enormous capitalisation costs (satellites, raw bandwidth, high-tech bandwidth-utilisation techniques), which must be amortised over use, even though the marginal cost of adding a new radio user is next to nil. This should be compared with the relatively high marginal cost of adding a new user to a wire system--the so-called "last mile" problem. In the decade time-frame, the choice for Universities seems to be the comparative cost of wires-- Ethernet/FDDI now or ISDN/ATM later. There is no longer a debate about where to lay them. Everywhere.
What do Users Want, and How do They Want it?
But how are these wires going to be used? One trend has been maintained throughout the history of computing. Users want: power; as much control over this power as they can get; and little responsibility for its upkeep. The ideal is to buy a small box, which does everything easily that one can think of, with which it is easy to communicate, which never breaks down, and which costs a pittance. The small box is here, the value of a `pittance' may be debated, some of them break a lot, all are confusing to fix, and you still need to be able to type. But modulo these details, many believe we are there already. New needs will develop, but computing life will only get easier.One consequence of this general trend is that students will become computerwise more independent, and independently more computer-wise. Here are some specific observations.
It seems more appropriate to go with the flow rather than against it. But is everything going to reside on private machines? What resides there now is general tools, that effectively use only the resouces on the desktop, and don't cost much money. Let's call this basic computing. There is certain computing power that Unis can provide for people that they currently can't provide for themselves, for example
These examples all have features complementary to those of basic computing, namely: specialist computing; resource hogs; high unit cost. Let's call this niche computing. The contrast suggests that Unis have a place in niche computing, even as basic computing moves away from centralised provision.
Niche Computing Could be Shared with Locals
One potential advantage for a Uni which concentrates on the niches is that these unusual resources could be shared with willing, and paying, members of the local economic community -- mainly industry, but especially small new companies who haven't been around long enough, or haven't ever had the resources, to capitalise this for themselves. For example, if a Uni buys capable parallel machines for physicists, it could sell or offer `spare' cycles to those who could not otherwise have access to those cycles for themselves. This community service is just as valuable in the long run as supporting the arts -- if only because many of those companies, if successful, probably will help to support local arts as well. Civic-mindedness is good public- and community relations. A similar point may be made about expensive, specialist software. Many Unis specifically exclude commercial activity from their computing support. But if the resources are there, why not share them? What is so different about supporting a local struggling start-up from other types of community service?The UbiquiNet
What other kinds of computing are there? Internet connections and use may very well become viewed separately from basic computing. Now, the Internet is accessed through basic computers, but cheap Internet-capable PDAs will be on the market soon, and PDAs are an idea that's been long coming, indicating that it won't go away soon either. Such PDAs are already almost there in the US. They'll have good text-processing and -saving facilities (for example, their file systems may become simply structured databases of text objects grabbed from the WWW, plus local DBs developed by the user) as well as very good mail-processing and WWW-surfing ability. That's significant work, but it doesn't encompass all of basic computing. It doesn't include research- or term-paper writing, doing general statistical calculations, writing small programs in specified languages for homework or research use -- in short, it doesn't include anything that my faculty, for example, currently wants a student to do! But it does encompass a need. Let's call it netting.Internet access devices share a lot of features with cellular phones -- in fact, some of them already incorporate cellular phones. However, it's hard to predict if cellular phones will remain tools for rich people, or whether they'll break the price barrier for general student communications devices, such as regular phones. One could guess that cellular will remain too expensive for student or general Uni use for five years, because of the amortisation of capital. `Cordless' may well be the lowbrow solution, as it is with phones. But `cordless' in the PDA world seems to mean infra-red connections to wires.
One may take the analogy further. PDA use may be much easier and cheaper through a service provider, just as, when phoning, it helps to have a phone company. If such providers develop, it is likely that they will be more efficient than a Uni trying to provide such services itself. Think of the plethora of products, protocols and services there is likely to be. Monopoly pricing, such as practiced now over most of Europe, could countermand this argument. But the days of monopoly pricing seem to be numbered -- Britain and the US don't practice it, and Germany and France will be changing within two years. It's hard to see prices remaining twice their real cost in the rest of Europe, and if the phone companies cannot retain a monopoly, it's hard to see how Internet service providers, which will use mostly the same lines, would be able to build one.
But, one might suggest, the Internet itself is a counterexample to this scenario --- a large, anarchical, but yet highly cooperative system based on highly detailed technical standards achieved by consensus. However, it must be remembered that the Internet is an outgrowth of academics, researchers, and committed hobby-nerds, and it is only comparatively recently in its history that industry and commerce has been hooked in. The PDA-scene is commercially-driven from the start, and where there's strong commercial competition, parallel academic enterprises have mostly failed. Except for Berkeley Unix.
Since most Unis don't provide phone service to students, I doubt they will provide this kind of Internet service either, at least not in this way. So, we must consider Internet service as separate from niche computing. And I have already argued that it's distinct from basic computing.
So Who Provides, and What?
Thus we can separate three different kinds of computing: PDA/Internet, basic computing (reading and writing papers, compiling and running programs, doing some calculations) and niche computing. Students and service providers will take care of the first, students will take care of the second, and it will be the Uni which takes care of the third. Supposing the analysis is fundamentally correct, I don't yet see any plausible scenario in which a Uni takes care of either of the first two.What infrastructure is required to support this? Much of it is probably in place. Many Unis have 10baseT lines everywhere, but not enough access points; similarly, there is ISDN capability on the phone lines, and a good bound for usage might be twice the number of current voice lines -- one can type on one machine and and talk on another at the same time -- and the phone comes through the machine. Student residences will have to choose between 10baseT-wiring and ISDN-capable phone lines to the Uni, which will be running high-bandwidth net access to outside. If PDAs come in through a service provider, the net will suffice. Basic computing may come in either through a service provider or through ISDN lines (providers are supporting Internet/PDA may not offer general windowing-telnet-type services and people may need to use phones with ATM). Niche computing will be used through whatever infrastructure supports basic computing.
Who Sets Trends?
Finally, how are styles formed? In many Universities, trends will be driven by whatever faculty and research groups are using, since they have significant leverage over Uni resources, and their resources are usually entirely Uni resources. Other than that, the low-end of the market drives trends outside the Uni as well as in.Summary
In summary, I have distinguished needs from styles, and clarified three potentially different categories of computing use: basic computing, niche computing, and netting. I discussed some social and economic factors driving the choice of categories as well as the styles of use, and predicted in general terms how these will be financed in the future.© P. B. Ladkin 1995. Quotation and use of parts of this note encouraged, provided source and author are explicitly acknowledged. In plain words, the usual convention. With the usual disclaimer-- these are personal views.
Some Comments from Others
Many thanks to those who have contributed to this discussion so far: Stefan Leue, Harold Thimbleby, Andreas Dieckmann and Jens-Peter Lindemann. Their comments follow.
[Professor Stefan Leue is in the Computer Engineering faculty of the University of Waterloo in Canada]
In general, I agree with your conclusions, but also see the potential conflicts with the particularities of the German Uni system.
Stefan Leue
[Professor Harold Thimbleby is Research Professor of Computer Science at the Middlesex University in England, and runs the University's WWW Server]
The distinction needs/style is useful, and can be refined further with "fashion." Uni students must learn to be discriminating, but are not necessarily so to begin with; their attention must be captured. Thus, our first year has lots of multimedia in it, and this requires CDs, speakers, colour ... stuff that is fashionable, but hardly necessary for Computer Science (unless you are specifically into multimedia as a CS issue, e.g., as real time databases) - in fact cute hardware as a medium for teaching, rather than simply content, is rarely necessary for any Uni course, except medicine, flying and a few other subjects.
Students can already buy (let's say) Nintendos that *appear* to be more powerful than Uni computers; they therefore expect this sort of cosmetic computing `power' here and/or they may select Unis on this superficial criterion. (So may their parents/teachers who may influence the students' entrance decisions.) I guess this issue will disappear in the future as PDAs (what we call *real* Nintendos) acquire connectivity etc and students can use their own material and software, at much lower cost to Unis. Unis may stay up-to-date fairly cheaply as most of the cost (and initiative!) would then be off-loaded.
We decided to buy a bunch of Toshiba notebooks that are managed by the Library. Students obtain them on loan, and so on. They have 10baseT Ethernet connectivity etc. Students have used them to give presentations in their course assessments.
We also run an advanced MSc in computer graphics as a way of (a) producing visible PR stuff, and (b) allowing kit to filter down to first years - though the latter idea doesn't work in practice as high end kit today is useless in 4 years time for many other purposes because of lack of support appropriate to the undergrad level.
Harold Thimbleby
Andreas Dieckmann, a student at the Universität Bielefeld, has pointed out that, no matter what happens in the future, every student should have the opportunity to follow the course of study that (s)he wants. Even though many students already have machines, there are those who aren't able to buy them yet. (This situation varies from country to country.) One should thus take care that future trends not dictate current requirements -- that while computers remain relatively expensive for some students, universities must still consider how to provide basic services, until such time as they are no longer needed.
Agreed. My essay says nothing about how to manage such transition. Only what form I think the transition will take. More worrying is the generalisation of Andreas's point to society as a whole. There is a danger, noted already by some US colleagues, that the increased use of computers for everyday life may lead to greater social inequity. For example, much of the Palo Alto city government's interactions with residents is now conducted via the WWW. Most residents of Palo Alto are computer-capable, so this is very effective. But suppose this form of political life extended to society as a whole. Those who are less capable of buying or using the requisite machines risk being excluded from `normal' political life. For students, this transition problem may be temporary, but for society as a whole, it may linger.
Peter Ladkin
Jens-Peter Lindemann, also a student at the Universität Bielefeld, and a system administrator for our faculty systems, noted that a transition to individual basic computing may be harder than it looks.
He notes that we should not underestimate the time and effort it takes correctly to configure a personal system for a particular network environment. Configuring e-mailers and other net systems is not a trivial task, and history (especially of the `reliable systems' such as Unix) has shown it becoming more complex, rather than easier. It may remain a tricky, technical issue for some time to come. This means that the resource investment needed to become effectively `wired' may remain large, independently of finances.
This is a very good point. Money isn't the only resource involved. Jens-Peter puts his finger on the very problem that is not obviously being solved in current system development. Were this `configuration' problem to remain difficult, this would reinforce the social inequality arising from computer use. Any transition planning must take this hard problem into account.
Peter Ladkin