Marathon Launches Fault Tolerance for Xen on Windows

Published: March 26, 2008

by Timothy Prickett Morgan

A year ago, Marathon Technologies, a maker of systems software that makes physical Windows platforms fault tolerant, said that it was working on a new line of products under the umbrella name v-Available to bring fault tolerance to virtual machines running Windows and, eventually, Linux. This week, Marathon is rolling out its first v-Available product, everRun VM.

The everRun brand name is not new, of course, and is in fact the brand used on the software that Marathon has been selling for years to make fault tolerant Windows boxes for companies who can't have Windows give them the blue screen of death for any reason. The VM part of the name is obviously short for virtual machine. Marathon has been around for 15 years and has close to 2,000 customers who have put its everRun products into their data centers to make their Windows boxes resilient. About a year ago, Marathon partnered with XenSource, the commercial entity behind the open source Xen hypervisor that was acquired by Citrix Systems last summer. The everRun VM product that is being launched this week is the result of a tight partnership between the two companies.

Unlike other high availability clustering solutions for virtual machine partitions on servers, everRun VM does not require a shared storage pool on a storage area network (all you need is a Gigabit Ethernet or faster link between machines) and it does not run the HA software up in the operating system layer above the virtual machine hypervisor. Because Marathon has a lot of experience in creating microcode for clustering that runs below the operating system, it was able to work with XenSource and then Citrix Systems to put the HA clustering software for fault tolerance as a software layer that runs between XenServer Enterprise Edition (which is itself running on a bare metal X64 server) and the virtual machine guests. By doing this, virtual machines and their applications and operating systems can be made fault tolerant without any changes to the operating system or the applications. This is a big deal, because fault tolerance in virtual environments becomes utterly invisible, just like it was in everRun configurations for physical servers.

And according to Michael Bilancieri, director of products at Marathon, virtualization is still in its early adoption phase, despite all of the hype, and that is because high availability and fault tolerance requirements that data centers have implemented for physical servers running mission critical applications cannot be virtualized until there is some sort of mechanism to ensure availability. "While people talk about virtualization being on between 6 percent and 10 percent of X64 servers being virtualized, a much smaller percent of these servers are actually in a production environment," explains Bilancieri. "That why our whole paradigm is to prevent outages before they occur, and this is resonating in the market." The reason is simple. Companies are wary of putting all of their eggs in one basket already, so cramming 20 virtual server images onto a single server gives IT managers nightmares because recovering from outages is 20 times more painful than if workloads were isolated on pairs of physical servers.

Right now, everRun VM is supported on the XenServer Enterprise Edition hypervisor, but because of the partnership between XenSource/Citrix and Microsoft, the virtual machine file formats for Xen and Hyper-V, the embedded hypervisor inside Windows Server 2008, will be compatible, which means everRun will be able to support Hyper-V virtual machines as well. Because of the closed nature of VMware's ESX Server hypervisor and the radically different file formats and mechanisms it uses, it is problematic to extend everRun VM to ESX Server.

Incidentally, basic failover for clustered servers running XenServer Enterprise Edition is expected to be part of an upcoming release from Citrix in the second or third quarter of this year. What Marathon is offering with everRun VM is not just failover and a best effort to move jobs from one hypervisor to another on two distinct machines, but a guaranteed failover, including data and I/O and provisioning for CPU, memory, and I/O resources necessary to run a job. Marathon will tweak everRun in the fourth quarter of this year to provide an embedded version of the everRun HA software to provide full lockstepping of application and memory state across virtual machines running inside hypervisor virtual machines. Bilancieri says that about 90 percent of the Windows applications out there need the component-level fault tolerance in the everRun VM product, which is why this one is coming out first.

everRun VM is in beta testing now and will be generally available in April; the software will cost $2,000 per server perpetual license, with a $500 per year software maintenance fee. Customers have to buy their own XenServer Enterprise Edition license, which Marathon will be able to resell. The software supports Windows Server 2003 and XenEnterprise today, and it will support Windows Server 2008 in the fall. Linux support, which was expected this year, might come out in 2009 now.

RELATED STORIES

Marathon to Bring Fault Tolerance to Linux in 2008

Marathon Makes Virtualization Fault Tolerant with v-Available

IBM Gives Its Seal of Approval to Key Information Systems' Innovation Center

Marathon Introduces High Availability Software for Windows

Marathon Technologies Targets Low-End Fault-Tolerant Windows Market with FTvirtual Server 6.1

                     Post this story to del.icio.us
               Post this story to Digg
    Post this story to Slashdot