Marathon extends HA software

By Ann Silverthorn

—This week, Marathon Technologies introduced EverRun HA, high-availability software that uses a virtualization architecture as an alternative to traditional clustering and fail-over. The software is built on the same platform as Marathon's EverRun FT (formerly FT Virtual Server) continuous availability software, but is designed for a wider market at a lower total cost of ownership (TCO).

Unlike Marathon's FT product, the two servers on which EverRun HA runs do not have to be identical. EverRun HA also adds multi-socket and multi-core CPU support for greater scalability.

EverRun HA creates a virtual application environment from two servers synchronized over Gigabit Ethernet. One of the servers is active and the other is "ready," but they both appear as one virtual server. Users can install applications to the virtual server with no need for modifications or scripts as in clustering.

Marathon's SplitSite option for EverRun HA allows the active and ready servers to be geographically separated by a distance of at least 100 miles depending on the type of connectivity in use.

"EverRun HA's differentiator is its policy management and failure detection as a plug-and-play solution rather than something users have to cobble together," says Gary Phillips, Marathon's president and CEO.

"In a cluster, the nodes have to be managed separately," adds Jerry Melnick, Marathon's vice president of engineering. "Users have to manage two copies of the operating system. They also have to write scripts to fail-over, and they have to write code to manage faults."

In contrast, according to Melnick, EverRun HA controls and protects the system without user involvement, since the application runs in the virtualized environment. By automating and embedding policies, the software eliminates human error in fail-over technology that often causes failure detection and fail-overs to be unsuccessful.

Phillips says if there's an outage in the disk, memory, or any I/O component on one of the servers, EverRun can "ComputeThru" the failure so there's no disruption to users. EverRun HA then automatically repairs the problem and resynchronizes the data without user intervention. For example, if one of the network adapters fails, EverRun HA allows operations to continue without affecting the applications' operations or clients' connectivity.

EverRun HA is priced at $7,500 for the single-socket, dual-core version, and $10,000 for the dual-socket version. EverRun FT is priced at $16,000.

This article was originally published on April 18, 2006