Client/Server: The Appliance of Clients
Client/server systems are proving a bit of a conundrum. Everyone agrees they will stay around, but in what form? Sean Hallahan investigates
It would require a latter-day Nostradamus to accurately predict the future of computing in five years? time, let alone 100. But an army of consultants, technologists, researchers, suppliers and writers have spent the last year attempting to evaluate the future of client/server technology.
Most observers are convinced that client/server systems are here to stay, but there is less certainty about the form they will take. There is the hardware debate about which machine is best suited to be the server, mainframe, mid-range system or high-end PC; the operating systems debate; the NT vs Unix vs proprietary and the database debate. But overshadowing them all is the question of distributed computing vs a return to centralised computing. Allied to this is the issue of which technology will dominate on the client side, the PC or the network computer (NC).
Until recently, the battle lines were clearly drawn. Those predicting a return to centralisation and the growth of the NC market included Oracle, Sun and IBM. Against them were ranged the forces of Microsoft, Compaq and others who had a vested interest in maintaining PC power. But within the last few months there has been the sound of bets being hedged. Microsoft and Compaq have admitted there is some validity in the NC argument while IBM has endorsed NT as an operating system on all its platforms.
US consultants the Aberdeen Group commented on IBM?s volte face in a report published in May: ?The political, religious and technological battles between Microsoft, with its NT operating system, and IBM with OS/2, AIX and MVS have ended and a surprise winner has been declared: the enterprise information technology customer ... this represents a major and long awaited shift in IBM?s previous anti-NT public position.?
In an extremely hard-hitting statement, IBM, while paying homage to NT, attacked it for its poor scalability and unsuitability for transaction processing. Microsoft?s Scalability Day, which took place on 20 May, was intended to demonstrate NT?s degree of scalability.
IBM rubbished the event before it had even taken place, a measure of its confidence that IBM alone could supply true enterprise-wide scalability with NT as the prime operating system. ?It is a good bet that much will be written and said about Scalability Day. That informal title is cause for concern because it would seem to divert attention away from the real issues and opportunities brought by NT. IBM, with its transaction and database processing software, is in a unique position within the industry to make that observation,? the statement said.
It went on to accuse Microsoft of taking only transaction throughput as a measure of performance. ?Simple read/ write transactions are a far cry from querying multiple heterogeneous databases and the other complex transactions that require many processing steps and that are often part of IT environments of the world?s largest corporations. In essence, the IBM case is that while NT is a powerful operating environment, it takes more than power to make an enterprise system.
IBM feels that a demonstration of theoretical scalability based only on transaction throughput does not reflect real life, and that NT is not up to the task without the assistance of other software. ?Multiple thousands of concurrent users executing multiple thousands of simultaneous transactions can cause systems stress, which in turn might result in software failures,? continued the IBM paper. ?The failures can be related to load dynamics and timing conflicts at the operating systems level, database, transaction manager, other middleware and sometimes in the application.
?Developing sophisticated load- handling techniques to avert these problems requires software that is robust, flexible and standards-based. NT has none of these characteristics. So far, a demo of two-node clustering is all Microsoft can offer in the area of fault tolerance.?
The statement went on to stress IBM?s commitment to supporting NT on all its platforms and operating systems and detailing its middleware offerings, the DB2 database, communications servers, Transaction Server, Lotus Domino Tivoli management systems and directory and security services that run on NT. It also points out that IBM?s services and support organisations are skilled in NT and support NT integration projects. ?In short, we believe that NT users are an important market for us and we are committed to being a leader in it,? IBM claimed.
Committed to NT IBM may be. Committed to Microsoft it is certainly not. ?But, NT and the related Microsoft Back Office products are not ready to meet the challenges of transaction processing as well as our customers? demand,? IBM added.
Phil Payne, a consultant with Sievers Consulting, is in no doubt that the future of client/ server lies with the NC or thin client terminal rather than the PC. Payne believes that enterprise-wide corporate IT will come more and more to resemble the utility companies.
?The more you centralise, the more you become like a utility company, like the gas or electricity suppliers,? he says. The choice of operating system is largely irrelevant, according to Payne, but good middleware is imperative. He does not think that Microsoft can supply the required middleware. ?Directory services has always been Microsoft?s weak point.?
Payne believes that IBM is planning to dump benchmarking as a means of assessing the processing power and speed of a machine. IBM uses a variety of benchmarks, including TPC, debit-credit and its own proprietary Ramp-C, to compare an IBM processor?s performance with that of a competitor. Occasionally it falls back on Mips (millions of instructions per second or, as some wags have suggested, meaningless indicator of processor performance). The problem with all benchmarks is that they often do not compare like with like.
IBM claimed it had audited debit-credit figures which showed that an IBM Mips is 3.24 times the size of a Digital Mips, according to the Handbook of IBM Terminology, published by consultants Xephon.
Now it seems that IBM is disenchanted with benchmarking, possibly because its machines are the standard to be tested against and partly because each time a rival?s benchmark shows an improvement on the IBM box it generates negative publicity.
In its Scalability Day statement, IBM hints broadly that benchmarks are no longer any real measure of power. ?With off-the-shelf generally available, DB2 and Transaction Service software, it took us just a few weeks to assemble a system to run a billion sample transactions in 19 hours in a networked NT environment.
However, this scenario doesn?t come close to simulating what happens in a real enterprise environment. Simple read/write transactions are a far cry from querying multiple heterogeneous databases that require many processing steps and that are often part of IT environments of the world?s largest corporations.?
The Aberdeen Group predicts that Compaq and HP will be IBM?s main competitors in the battle for the NT-based client/server environment. It thinks that while Compaq will make an impact at the low end of the market, it will fail to penetrate the high end. ?Compaq does not have a worldwide service and support organisation in place to help address enterprise level sales, service and support requirements ? and Aberdeen suspects that it would take Compaq at least five years to assemble a global support infrastructure capable of effectively competing with IBM,? the group?s report says.
The Aberdeen Group believes HP is in a better position to challenge IBM for the high-end NT market but points out that the two companies? strategies differ markedly. HP?s strategy is to meld together its Unix offering with similar products from Microsoft. ?Applications unity, linking existing HP applications with Microsoft applications, is the central theme at Hewlett Packard,? the report says. IBM?s strategy is directly the opposite.
?IBM?s strategy, on the other hand, is to de-emphasise the melding of IBM applications with Microsoft applications ? IBM is clearly showing that it believes that it can deliver enterprise strength, robust applications and middleware on a more timely basis (and with greater functionality) without having to merge applications and middleware developments with Microsoft,? the report says.
IBM is convinced that the NC will play an increasingly important role in the future of client/ server, according to Tony Occleshaw, marketing manager for IBM?s European software business. He concedes that there will still be a role for the PC but argues that end users are only now discovering the true cost of ownership of PC networks.
He also admits that many customers are turning increasingly towards NT and maintains that IBM is fully committed to meeting its customers? requirements. ?The big change came two years ago when The IBM Software Group became an independent business division. We are now platform-agnostic,? he says.
Occleshaw thinks IBM can deliver a better software solution based on NT than Microsoft can, because of IBM?s strong middleware, data integrity and data security products. He believes the internet will play an important role in the future of client/server but will be slower to take off than many pundits have predicted.
Whether or not the NC takes over almost completely from the PC Lan in the corporate environment is still a matter for some debate. Although Microsoft has endorsed the concept of the thin client, it still has more to lose than any of the other players if the NC takes off. The Web browser also presents a threat to the company, according to Bloor Research. In its report, The Enterprise By Other Means, Bloor examines the future of Microsoft operating systems.
?Windows had become the widely accepted standard and the assumption had been that Microsoft could control its evolution indefinitely, from Windows to Windows 95 to WNT,? says the report.
?The proliferation of browsers has broken Microsoft?s grip in two ways. The Windows look and feel can no longer be controlled. The browser is the user interface. Java executes within the browser and will gradually eat the PC apps one by one. As far as the user is concerned, the browser will become the operating system.?
The last great ?revolution? in computing was the birth of the PC in 1981. In the late 1970s, very few companies foresaw the impact the PC would have on corporate computing.
This time around, everyone is thinking ahead and speculating on the future of client/server. Picking the future winners and losers is no easy matter, even for the industry consultants. But of one thing there is no doubt ? the future of client/ server computing is undergoing its greatest change ever and winners and losers there will be. Whether the IBM, Sun, Oracle camp or the Microsoft, Intel and Compaq bloc comes out in the ascendency will doubtless become clear shortly. One other thing is also clear ? not everyone can be a winner.