Microsoft Corp. plans to use Marathon Technologies' fault tolerance expertise to bolster Windows Server and, eventually,...
Marathon's everRun fault-tolerance software enables virtual machines (VMs) and Windows applications to run in lockstep on two servers. When hardware on one server fails, failover software enables applications to keep running.
As IT shops delve further into virtualization for mission-critical applications, they have turned to failover software such as everRun from the Littleton, Mass.-based Marathon to reduce application downtime.
"We are seeing more people relying on Windows Server to host mission-critical apps, beginning with Windows Server 2000," said Mike Schutz, the director of product management Windows Server Division at Microsoft. "With server consolidation and virtualization, the impact of downtime is even greater than ever."
But that doesn't mean every IT shop needs the same level of failover – or that it needs failover software at all, virtualization analysts emphasize. It depends on the kind of applications a company runs.
How everRun failover software works
EverRun currently works with Windows Server 2003, and by Q2 of 2009, it will support Windows Server 2008. Last week the two companies said they will will also bring everRun technology to a version of Hyper-V following the Windows Server 2008 R2 release, now in beta, said Schutz. The release date has not been disclosed.
In the meantime, Windows Server 2008 and Hyper-V users can use Windows' fail-over clustering capabilities.
Also, for customers requiring a higher level of availability, Marathon everRun is already available for Citrix Systems' Inc.'s XenServer, which supports the most recent Windows versions. Under that model, everRun VM software is installed on top of a hypervisor, and a Web-based management console that allows users to view VMs and choose which VMs to protect. The software protects VMs by creating redundant VMs and synchronized mirroring of the network, storage and data. The software works with any Windows Server application. Ultimately, Microsoft's and Marathon's goal is to offer joint customers the ability to protect applications with different levels of failover.
Level 1 offers basic failover clustering, as provided by the Enterprise and Datacenter Editions of Windows Server 2008. Level 2 provides component-level fault tolerance, which protects against the failure of an individual CPU or network card, for example; the list price is $2,000 per server and covers unlimited VMs. Finally, Level 3 offers system-level fault tolerance and protects against any kind of failure by maintaining applications and memory in their pre-failure state. This level will begin shipping in April, and will have an estimated price of about $5,000 per server, according to Marathon.
Does every shop need fault tolerance?
But while virtualization has made fault tolerance software more pervasive, advances have also made the technology easier to obtain. Thus, using fault toleranc eno longer requires specialized IT skills and expensive hardware, because everRun software is completely automated, said Gary Phillips, president and CEO of Marathon Technologies.
"We clearly see the trend to extend availability, and customers want something simple and automated, which our product provides," Phillips said. "Admins can extend their high-availability capabilities in Windows Server 2003 and, later with Windows Server 2008, with fault tolerance-level availability in minutes, without any specialized IT skills."
Further, true fault tolerance is a niche requirement for some IT shops, but not all, contended Andi Mann, a research director at Enterprise Management Associates.
"Not every organization needs true fault tolerance," he said. "But for those that do, this functionality is absolutely critical. The question is whether you can tolerate the downtime when -- not if -- it occurs. For most organizations, even a few minutes a day is acceptable, albeit not ideal. … [But] in very high-value applications like trading-floor systems, a minute of downtime can cost millions; in a hospital or military application, a minute of downtime can cost lives. In these scenarios, unexpected downtime is completely unacceptable."
Building vs. buying virtualization capabilities: The need to compete
Microsoft needed to add Fault tolerance capabilities to Hyper-V to better compete with VMware and Citrix in niche markets, Mann argued.
While Citrix and Microsoft now use Marathon's technology for virtual machine (VM) fault tolerance, VMware has developed its own flavor, which it introduced in September at VMworld 2008. It is due out sometime this year as part of VMware's Virtual Datacenter Operating System (VDC-OS).
Schutz said Microsoft chose not to develop its own fault-tolerance software because Marathon is already a partner and has built products on and around Windows for years. "It should be no surprise that we chose Marathon; they have been a leader in this area and have a tremendous amount of expertise. It was a natural partnership for Microsoft," Schutz said.