Microsoft misses performance and scale goals with Viridian – Microsoft
One month trip from gloating to tears
Viridian has gone on the Redmond Diet with Microsoft today ripping some of its most exciting planned features out of the virtualization software.
In April, Microsoft's GM in charge of Viridian Mike Neil revealed that the company would have to delay the software's beta release from the first half of 2007 to the second half. The reason for the delay? Well, Microsoft wanted to add in things such as support for 64 processors – "something no other vendor's product supports" – and on-the-fly addition of processors, memory, disk and networking. Such technology was needed so that Microsoft could "(meet) our internal goals for performance and scalability."
Consider those goals not met as of today with Microsoft saying it will only support 16 processor cores in the final version of Viridian – or more formally Windows Server virtualization. In addition, Microsoft will give up on the on-the-fly technology and give up on live migration with the virtualization software release.
Neil disclosed these changes in a blog post under the heading "Dynamic datacenter."
Customers will think Microsoft's data center dynamic in all the wrong ways. Microsoft has now fallen even farther behind rivals VMware and Xen in the server virtualization game.
Both competitors support on-the-fly additions of computing resources and have tools for so-called live migration where you can shift running virtual machines between physical servers. Such tools are thought key to the oncoming wave of data center-wide virtualization software hitting the market.
The 64-core loss hurts few since most x86 machines top out at four sockets – or 16 cores with four-core chips. Still, it's rather hilarious that Microsoft had to give up on that feature since Neil was boasting about crushing rivals with the SMP play last month. (Someone pull Neil away from his blog – Ed)
This time around Neil confessed that Microsoft's bizarro version of feature creep comes from the realization that "shipping is a feature, too."
"So we had some really tough decisions to make," Neil wrote today. "We adjusted the feature set of Windows Server virtualization so that we can deliver a compelling solution for core virtualization scenarios while holding true to desired timelines. Windows Server virtualization is a core OS technology for the future, and we chose to focus on virtualization scenarios that meet the demands of the broad market – enterprise, large organizations, and mid-market customers."
The rest of Neil's blog post proves worth reading as it details some of the features that will actually
allegedly ship with Viridian, which is due within 180 days after Longhorn Server ships later this year.
Neil also expressed optimism about the sever virtualization market still being so young. Less than 10 per cent of customers have picked their VM vendor of choice.
That's a fine thing to bring up if you're VMware or Xen with modern product on the market. Instead, Microsoft remains locked in the hosted hypervisor world where its current Virtual Server product has to run on top of an operating system. And now its already delayed software will arrive missing features. You're basically looking at a first generation virtualizatoin product that will ship, say, a year from now.
These issues have some analysts calling for blood.
"Microsoft has a fundamentally broken server virtualization strategy at this point," wrote Illuminata analyst Gordon Haff. "They were behind to begin with. Now, the tardy 'Rev. 1.0' is starting to look more like 'Rev. 0.5.' Perhaps it’s time for Microsoft to consider a different angle.
"Perhaps it’s time for Microsoft to admit that they can’t do it all themselves - at least for now - and form some legitimate partnerships. That would mean fixing some licensing problems and eating some crow. But that’s the cost of a broken internal development process. That Mike Neil should make reference to the 'mythical man-month' in his posting is wholly appropriate." ®