Is Microsoft secretly preparing for a new computing scenario where a reborn mainframe emerges that supports multi-user computing on centralized desktops?
Directions on Microsoft analyst Rob Helm believes so, pointing out that two of Microsoft's most recent acquisitions-Calista Technologies and Kidaro-have been companies whose technologies could facilitate multi-user computing on centralized desktops.
Helm was responding to a Reuters report published March 13 quoting Microsoft's chief research and strategy officer, Craig Mundie, as saying that he believes parallel computing is the next big thing in technology.
But individual parallel applications might be the wrong way to exploit parallel processors, Helm said, noting that a better way to exploit parallelism could be to "go back to the future," essentially sharing computers among multiple users.
"A reborn mainframe could spell the end of the PC market, so Microsoft and Intel aren't going to talk about it out loud, but I think Microsoft, at least, is quietly preparing for the possibility," Helm told eWeek.
Online services, another big future growth area for Microsoft and one of the reasons behind its drive to buy Yahoo, are also inherently multi-user systems. Such services will be able to exploit continued improvements in parallel computing. "Google, of course, has built its whole business on that observation," helm said.
Mundie is not alone in this assesment, with Hewlett-Packard and Cisco Systems also saying that parallel processing would be the big driver for the data center.
A new programming language would be required for this new parallel computing scenario, and could affect how almost every piece of software was written. "This will be hard," said Mundie, who worked on parallel computing as the head of supercomputer company Alliant Computer Systems before joining Microsoft. "This challenge looms large over the next 5 to 10 years," he told Reuters.
But Helm says that while exploiting parallelism does require a fundamental shift for programmers, he does not believe that the tools and languages they are given today are up to the task.
"However, new tools and languages might not be sufficient either. For years, researchers have been experimenting with new programming technology to support parallel computing in scientific labs, but very little of it has made the jump to the mainstream," he told eWeek.