By Lisa Coleman
This month, San Jose-based Kashya began shipments of a data replication product that includes a full range of replication options for linking heterogeneous storage systems over long distances.
In the past, end users were often limited to expensive storage-specific (proprietary) data replication options from major vendors. But a number of start-ups are introducing lower-cost replication products that can work with virtually any vendor's disk arrays.
Kashya's KBX4000 is an appliance that enables replication between heterogeneous disk subsystems. The appliance sits in the intersection of a storage area network (SAN) and a wide area network (WAN) between primary and secondary sites (see diagram). It offers any type of data replication—synchronous, asynchronous, point-in-time snapshots, and "short aperture" snapshots (e.g., snapshots taken every second)—with no distance limitations or application degradation, according to company claims.
Kashya is positioning the KBX4000 as a replacement for both array-based software (such as EMC's SRDF and IBM's PPRC) and host-based replication software (such as Veritas' VVR).
"The value that companies like Kashya bring to the party is that they don't care what storage is at location A or location B, and they give you give a variety of replication options such as synchronous, asynchronous, and point-in-time," says Arun Taneja, founder of The Taneja Group consulting firm.
"The problem with what's out there today is that for every level of protection there's a different solution," says Mehran Hadipour, vice president of marketing at Kashya.
Kashya's KBX4000 appliance sits at the junction of a SAN and WAN and provides any type of data replication.
The KBX4000 can support any type of storage, server, or fabric, and replication is policy-based with users selecting the number of volumes and method of replication. The appliance uses data compression for more-efficient bandwidth use and lower cost. For example, it reduces bandwidth requirements by maintaining write-order fidelity, checking for multiple writes to the same address, and transferring the most recent write only for each block. In addition, the appliance has a native WAN interface, which eliminates the need for costly edge connect devices such as routers and gateways.
Software runs on two pairs of IBM xServers. Pricing is based on the amount of data replicated and starts at about $65,000 for 1TB ($50,000 for the software license and $15,000 for the two pairs of servers), according to Hadipour. The company recommends using one pair of servers on each side (primary and secondary site) of the infrastructure for fail-over.
The appliance uses standard IP protocols for moving data bi-directionally to enable two sites to backup each other, says Hadipour.
"The challenge for these new players is to guarantee recovery in a consistent state on the remote side," says Taneja. To achieve this, Kashya time-stamps changes before they are sent over the wire so that the devices can essentially "talk" to each other.
Other companies working on similar technology include Avamar and Topio. Separately, EMC recently announced a lower-cost version of its SRDF replication software. SRDF/A (asynchronous) lets users maintain remote copies of their data at any distance with no host application impact, according to company claims. SRDF/A is priced from $20,000, with cost based on capacity tiers.