Boxboro, Massachusetts-based Marathon Technologies Corp. has introduced an enhanced version of its high-availability kit for Windows NT servers, featuring new technology for tying dual-processor Intel Corp. servers into near fault-tolerant configurations.
Marathon’s new Multiprocessor Endurance 6200 hardware and software technology will also allow Windows NT users to increase the geographical distance between their servers to 500 meters, providing better disaster tolerance. The tool kit is priced at less than US$50,000 per kit.
Marathon Technologies has been trying to carve a niche for itself selling a patented technology called ComputeThru. The company claims ComputeThru delivers 99.999 percent uptime – or less than five minutes of downtime annually – on Windows NT server hardware.
Hewlett-Packard Co. recently announced models of its NetServer systems featuring Marathon’s new technology. HP has a reseller agreement with Marathon.
At a basic level, Marathon’s approach involves tying four Intel servers together into one large, logical server array. The array is split into halves, with each half running the same application identically and in lockstep with the other half. If one half of the array fails, the application keeps running on the other side, said Craig Jon Anderson, a director at Marathon.
The approach ensures that there is no single point of hardware failure, Anderson added.
It also ensures that there is no service disruption in the event of a hardware failure, said Denis Nothern, information technology director at Mothernature.com Inc., a Concord, Massachusetts-based online retailer.
“The fact that the (array) can continuously run and be fixed without taking the whole platform down is remarkable” when compared with other approaches, such as Microsoft’s Windows NT clustering technology, Nothern said.
It’s the same reason First Options of Chicago Inc. is using Marathon’s technology to run an important trading application, said Brian Slattery, manager of server support. First Options is a provider of clearance and execution services to traders on major U.S. exchanges.
Unlike other high-availability techniques such as clustering, Marathon’s approach doesn’t allow for even a temporary service disruption if a server goes down, Slattery said. “That’s critical for us. A fail-over of even a minute or two could cause big problems” when dealing with trading applications.
“Marathon is the only vendor offering five nines of system availability in the NT industry at this time,” said Joe Clabby, an analyst at Aberdeen Group Inc. in Boston, referring to 99.999 percent uptime.
The new ComputeThru technology in Marathon’s tool kit can be used to tie more scalable and powerful dual-processor Intel servers into similar configurations.