Slashdot Mirror
Microsoft to Storm Linux Strongholds
VitaminB52 writes "Microsoft is only winning about one out of four deals where IT shops are trying to move off of proprietary Unix. To turn that trend around, there are four specific Linux strongholds where Microsoft is focusing its attention." From the article: "After discussing server clustering, Web hosting, and server appliances, Ballmer was cut off by the interviewees before he could identify the fourth. But my guess is that, given the way Ballmer emphasized Software as a Service (SaaS) as a core theme for all the work that's taking place at Microsoft right now, the fourth stronghold of Linux that Microsoft wants is the SaaS stronghold where Linux is the operating system behind a Java-based application server technology ... Ballmer knows he's got a long roe to hoe. 'The day I come in front of the Gartner audience and say we have a better Unix than Linux, that'll be a good day.'"
...the fourth stronghold of Linux that Microsoft wants is the SaaS stronghold where Linux is the operating system behind a Java-based application server technology
Sure, that makes sense, especially considering the big announcement last month of JBoss partnering with Microsoft to build up interoperability with Windows servers and the JEMS stuff.
Actually, it's already happened. Sun is getting widely dumped in favor of RedHat/x86, who is seen as an top-tier enterprise vendor in Sun's traditional strongholds. As a result they've been forced to adopt a very aggressive x64 strategy.
Business. Numbers. Money. People. Computer World.
"long roe to hoe" that should say "long roW to hoe". It's a term taken from agriculture. So sad that people are so removed from agriculture now that they don't what the term means or it would have been spelled correctly instead of after Roe v. Wade, the most common way now that people see that sounding word.
Microsoft and Ballmer just don't get clustering at all and I feel sorry for the 25% that got sucked in by M$ BS. Ballmer is bringing spoons to a steak party.
An OS that is graphical wastes resources in a clustered environment. It wastes CPU in managing it; it wastes electricity in powering it and adds to the total BTU output that raises A/C costs. Forget about the complexities added in that M$ solutions are new, poorly tested and of beta quality when compared to any UNIX/POSIX type OS. None of the aforementioned adds value to the compute task and often detracts from it. Most can be critical project problems if not managed and planned for.
One also has to look at the software acquisition economics. Say you have a 1024 node cluster. 1000 * 1024 for server licenses is $1M $$. FC4 is out and even if you used commercial Linux you would never pay $1M for this quantity of licenses unless they tossed in the installation and configured the cluster for you.
There are also other issues such as kernel/network performance and tuning but I will skip this.
My dream cluster would be few thousand Linux AMD 64 dual core, dual CPU systems with 16GB of ram in a 2 or 4 U package with front loading drives and can be managed without a VGA... hm... this OS/hardware exists without Microsoft!
Doug - a genuine Cluster Monkey
HPC for Primates. Read Cluster Monkey