Storage start-up Kashya this month is expected to roll out a server appliance that lets users replicate data over any geographic distance easily, affordably and automatically.

The company, located in San Jose and Israel, is named after the Aramaic word kashya - meaning a puzzling problem. Its KBX4000 WAN storage appliance sits in the network path between a storage-area network and a WAN router, automatically adjusts the bandwidth that will be used and replicates data between geographically separated SANs using IP.

Three researchers with the elite Israeli Defense Forces founded the company in December 2002.

Kashya's product solves how to replicate business-critical and non-critical data across varying geographic distances at speeds that let the data be sufficiently and affordably protected.

"The Kashya appliance is the first product I have seen that can modulate the bandwidth and pre-cache the data that is going across the wire," says George Crump, a manager with storage integrator Sanz in Castle Rock, Colo. Sanz is evaluating the box.

"That way, you can set policies by the criticality of the application and save money because we don't have to be paying to replicate all applications in the same manner," Crump says. "In our test environment, we are replicating [business-critical data] across a [slow] less-than-fractional T-1, and it's performing really well."

Replication with the KBX4000 is managed by rules-based software of Kashya's design that lets administrators look at their data needs and allocate bandwidth to the application. Policies set by the administrator are enforced automatically.

For instance, take a company that has a business-critical financial system that it wants to replicate as quickly as possible and at the longest distance possible with the least possible lag time between updates. The administrator would set the policy to synchronous replication, in which each data transfer requires an acknowledgement, and the lag limit, which specifies the amount of time to wait before transmitting more data.

Asynchronous replication would be set when the user wants to transmit data between more distant sites and as quickly as possible or when less business-critical data needs to be transmitted. If the system determines it can do point-in-time copies a few seconds behind, it will revert automatically to that method.