The Kashya KBX4000 disaster recovery system is an interesting product that can help IT managers replicate their SAN data over a WAN.
eWEEK Labs recently went to Kashya Inc.s San Jose, Calif., offices to get a look at the Kashya KBX4000 in action. We believe that the system, which is slated to ship this month, will be helpful to midsize businesses in meeting their data protection needs. The Kashya KBX4000 has numerous compelling features to optimize data replication while keeping an eye on bandwidth utilization.
An edge device, the Kashya KBX4000 sits between the storage area network and the WAN and acts as a gateway between geographically separated sites.
As write transactions are sent from servers to their respective storage systems, the Kashya KBX4000 mirrors those transactions to a local appliance and processes this data so that it can be efficiently sent to a sister appliance over the WAN.
The Kashya KBX4000 requires the installation of a filter driver on each of the protected servers. The filter driver mirrors the write commands sent to storage systems and sends them to the local Kashya KBX4000 appliance.
Given the current hellish state of patch management and the fact that additional software consumes valuable CPU cycles and memory, most IT managers see the addition of agents and drivers as a negative. As such, the need for a filter driver could be a factor in deciding whether to implement this product.
Kashya is addressing this potential problem by working together with switch vendors so that the mirroring process can be handled in the switch itself, instead of on each server. This is a smart and proactive step, and one that we wish more vendors would take.
Heterogeneous hardware support is a strength of the Kashya KBX4000 because it will allow mirroring between virtually any storage systems. (In contrast, most storage hardware vendors want you to mirror data only to their own products.) The benefit is that you can mirror an expensive Fibre Channel storage system in your primary site to a less expensive solution based on SCSI or ATA drives in a secondary site.
The Kashya KBX4000 sends data over the WAN using a couple of different technologies, including snapshots and asynchronous mirroring.
Kashya uses a rapid snapshot technology to capture transactions and send them over the wire. The advantage of doing it this way instead of just streaming out transactions is that the snapshots provide point-in-time restoration. In the event of a data corruption incident, the secondary site can be rolled back to a point before the corruption.
Before putting data on the wire, the Kashya KBX4000 uses compression and analyzes data to make sure that only data deltas are sent over the wire; this keeps bandwidth usage in check and allows IT managers to set up mirrors on smaller WAN pipes.
The Kashya KBX4000 management interface allows IT managers to set and enforce data protection policies. For example, we could set the limit on the amount of data queued up to go over the WAN.
This is extremely useful because it allows users to fine-tune policies to suit their applications. For example, a point-of-sale application server (where transactions are critical to business) can be forced to send data over the WAN immediately, while a less critical application such as e-mail can be made to wait a little longer before sending data.
The bandwidth optimization can be set according to a couple of different metrics, including number of transactions and the amount of data that needs to be sent.
Using SCSI commands, the Kashya KBX4000 can tell servers to slow their transactions to ensure that data crosses over the WAN before more transactions are processed. Although this is useful, we recommend that IT managers test it thoroughly before implementing it because the artificial latency introduced may not work well with some applications.
Bidirectional data transfer is not a problem for the Kashya KBX4000; if an IT manager has two active sites, mirrors can be sent to both sites. (For example, site A sends data to site B, while B sends data to A.)
Kashya officials highly recommend running redundant appliances at each site.
A configuration with four appliances (two on each side) and software will cost about $30,000 to protect 250GB of data. The appliance can scale to handle as much as 10 terabytes of data, but organizations will need to expand their licenses to handle larger loads. ´
Senior Analyst Henry Baltazar can be contacted at [email protected]