Microsoft's days are numbered. So are Intel's. And the rest of the PC industry's. That was the gloomy contained in a damning report from Bloor Research - The enterprise by other means - published last month.
The document raised so much concern - and so many hackles - that Bloor invited speakers from Microsoft, IBM, Oracle, Sun and Wyse to discuss them at the Commonwealth Institute in London last week.
Robin Bloor, chairman and chief executive officer at Bloor, opened with the failings of client-server computing. According to a Standish report, 47% of client-server projects are never implemented. And, in those applications that are completed, between 50% and 80% of programming effort is devoted to the issue of communications in a client-server environment.
Strengthening his case against client-server computing, Bloor quoted the 1996 Aspect International Computing report which found that 37% of applications in a client-server environment fail to scale up to the enterprise.
Projects generally exceed budgets, Bloor stated. Development time is slow and problematic, especially in larger systems. Worse, Bloor said that users do not really understand the complexity of client-server development in a distributed computing environment.
"The PC was not designed as a client machine; Windows was not designed as a client operating system; PC applications were not designed as client applications," said Bloor. Bluntly, he feels the PC is simply not suitable for distributed computing.
The weaknesses he has identified in the PC architecture relate to just about everything you could care to mention; security, performance tuning, consistency, integrity, recovery, back-up, availability, deployment, management problems, maintainability, support and archiving.
These attributes of a distributed computing environment do not apply to the system as a whole on the PC. For instance, Bloor pointed out that the rate of failure of PC LAN servers is completely different to the failure rate of an enterprise server.
Bloor looked at network computing and how this overcomes the limitations in client-server architectures. An immediate benefit to moving to network computing, he explained, is that it removes a layer of complexity.
In other words, all management issues on the client machine evaporate.
Management of the desktop infrastructure occurs on the LAN server. Within a year, Bloor predicts the removal of another tier of complexity, leading to a time where an NC device talks directly to a server.
Regarding the dependency of PCs on specific operating systems, Bloor said the days of the proprietary operating system were over. "We never buy operating systems for features. Instead, we buy applications; we buy benefits." Java, explained Bloor, "cuts the string between operating system and benefits."
It's almost a year to the day since Bill Gates said the Internet was not worth Microsoft dedicating a division to. In the intervening 12 months the company has spent over $2.1 billion (#1.25 billion) on NC and PC research and development.
The Net has evolved independently of operating system or hardware-specific platforms. It is now at a stage where people can run serious applications within a Web browser. They don't even need Windows. So Microsoft takes the threat of the Internet to its operating system and desktop applications monolopy very seriously indeed.
Jeremy Gittins, Microsoft UK's marketing manager for Internet tools and platforms, did not deny the need for NCs, but disputed their suitability as general purpose corporate desktop machines.
Quoting a CTI report, Gittins said only 10% of employees perform task-based roles in their employment. The other 90% are knowledge-based workers who, presumably, would require PCs, with PC operating systems and PC applications.
It is the 10%ers, the people who can live with a single application, that the NC targets, he claimed. According to Gittins, the bulk of corporate users either require mobile computers or local data storage, both of which are missing from the NC.
He argued NCs were not compatible with standard PC architectures. "They are not even compatible with each other." He added that because the NC Reference Platform doesn't specify a processor architecture, "people will have to support different hardware (platforms)".
Turning the cost of ownership argument on its head, Gittins claimed NCs were, in fact, significantly more expensive because they were not scaleable.
He contended that because an NC is independent of both the server and the network, it is necessary to increase network bandwidth in order to scale it. This can work out extremely expensive. But, what he failed to mention was the fact that the same is true for client-server computing, an area where Microsoft is making big in-roads. On the 'fat' client PC architecture, data is constantly being sent back and forth over the network to be processed on the client PC machine. This can quickly eat up available bandwidth, slowing network throughput and responsiveness.
On the server, Gittins again seemed at odds with other Microsoft activities.
This time it was the operating system message. Gittins took the "slice price" case, in which a massively parallel multi-processor server works out more costly than the equivalent number of desktop PCs, when counting numbers of processors. But there is every indication that the NT team at Microsoft wants the operating system to scale to symmetric multiprocessing architectures.
Gittins also spoke about Java. "Is it as strong as an operating system or a chip?" he asked rhetorically, questioning the stability of a programming language which has only been available since May. Gittins said that having one language for the Internet would severely restrict choice. "(Users) will have to throw out old code and their expertise," he added.
This does not correspond with IBM's version of events. IBM has put Java everywhere: on its System 390s, AS/400s and RS/6000s, so customers can keep their existing applications, Java-enable them or write completely new Java applications.
David McAughtry, general manager of the network computer business unit for IBM's EMEA territories, reflected on the history of the PC and how today's machines are 1,000 times more powerful, in terms of processing power. However, he felt PC applications had not improved 1,000-fold.
"You can put more and more power into a PC without producing a dramatic improvement. It does not necessarily improve things for your business.
If PCs were cars, they should be capable of a million miles an hour by now," he said.
Tony Occleshaw, competition marketing managing consultant at IBM, and a delegate to the debate, was dismissive about Microsoft's stance on network computing: "Microsoft has completely missed the point of the thin client.
The hardware is completely irrelevant."
Regarding Gittins' remarks on the need for local storage to satisfy the needs of knowledge workers, Occleshaw added: "I fail to see how a thin client cannot be used to access data. Microsoft is attempting to legitimise the need for a real hard disk."
Andy Bailey, product marketing manager at Oracle, claimed Microsoft has been fighting an on-going battle for control of the Internet over the past few months whereas Oracle's mission was "to make sure the Net remains open".
Bailey said it wasn't in anyone's interest to own the Internet: "Openness leads to competition which leads to good innovation."
If nothing else, Bloor's report appears to have put Microsoft on the defensive. Much the same can be said for network computing, which offers a viable alternative to the PC-centric way of working that has prevailed for 16 years.
In fear of future shortage - or in preparation for its own electric car project?
But if you're running anything older you'll have to wait
Powered by servers based on Qualcomm's scalable 48-core Centriq 2400 10nm CPUs
Malware has been in circulation for more than a year