Marathon Ships High Availability Virtualization Solution

Marathon Technologies announced Monday it is shipping a new virtualization solution that provides high availability for applications at a lower price point than traditional clustering. On top of that, the company claims, it’s easier and less expensive than traditional clusters to deploy and manage.

EverRun HA, from Littleton, Mass.-based Marathon, provides what the company refers to as a “pre-mixed” or turnkey high availability solution as opposed to the “roll-your-own” model of most clustering solutions.

To date, virtualization has found a growing audience in the area of infrastructure management with products such as VMware’s ESX Server and Microsoft’s Virtual Server. Marathon’s use of virtualization, on the other hand, is for a different purpose.

“We’re using virtualization to provide high availability [for Windows server applications],” said Gary R. Phillips, president and CEO of Marathon.

Specifically, EverRun HA virtualizes the application’s input/output (I/O) using Marathon’s patented “ComputeThru” technology. The application and operating system behave as if they were running directly on the hardware, but actually everything is intercepted at the I/O level.

EverRun HA uses a pair of x86 servers both running the same software to eliminate single points of failure. The software shields the application and the redundant platform by inserting a virtualization layer. It creates a virtual Windows server environment where the application is installed, operated and managed. The whole system appears to be a single standalone server, with a single identity and IP address.

The protected application runs on one server at a time, using the redundant hardware in the second server in the event of a device failure. This also means that the operating system level does not have to deal with issues like whether or not a device driver is up to date or not.

Because EverRun HA captures everything that occurs at the I/O level, a failure is transparent to the application. Marathon’s software immediately redirects the I/O – that is, reroutes the activity – to the secondary server so nothing is lost and the user experiences no interruption.

If a failure occurs that completely takes down the primary server, there will be a short interruption of service while the secondary server – or in Marathon’s lexicon, the “co-server” – and the application, restart. Still, nothing is lost, Phillips says.

Another benefit: clustered applications typically need to have customized operating system-level scripts, which need to be written, tested and maintained, in order to set up applications to handle automated failover. Applications do not need to be “cluster aware.” Neither does there need to be any modification of the application or the infrastructure in order to install EverRun HA.

Additionally, EverRun HA does not require that both servers have identical hardware.

Ultimately, Marathon’s sales pitch is high availability without the hassles of clustering at a significantly lower price than a clustered solution. EverRun HA starts at $7,500 and comes with a year of premier support. It runs on Windows Server 2003. Users only need a single application license.

EverRun HA requires two standard x86-based servers tied together by two gigabit Ethernet links, and it supports direct attached storage, network attached storage and storage area networks.

About the Author

Stuart J. Johnston has covered technology, especially Microsoft, since February 1988 for InfoWorld, Computerworld, Information Week, and PC World, as well as for Enterprise Developer, XML & Web Services, and .NET magazines.