XIOtech brings clustering to SANs

This Monday XIOtech, previously regarded as a midrange player when it comes to computer room storage, stepped up to the plate with a product squarely aimed at enterprise storage environments.  Its Magnitude 3D line of storage servers brings clustering technology, previously available on application servers, onto storage-area networks.

Clustering of course is nothing new, and has been around since at least the mid-1980s (remember VAX clusters?). Certainly many of today’s blade servers are a clustered architecture in a single rack. But XIOtech seems to be the first storage company to put clustered storage on a SAN, and to allow systems within the cluster to be dispersed up to 300 meters apart.

What is clustering, and what are its benefits? 

As you might expect, clustering can be implemented in several ways, but it is useful to think of clustered servers as being two or more devices that work together so they appear to the user as if they were a single system.  Multiple redundant connections run between the various nodes of such a system, carrying both application data and a monitoring “heartbeat.”  This heartbeat signal runs across the cluster several times each second. 

When one of the devices within the cluster fails to send out a heartbeat that can be read by the other nodes (or, in some implementations, by a designated master node), the system assumes something is wrong and failsover the functions of the failed node to another device within the system. When the heartbeat reappears after the problem has been fixed, processing can be transferred back to the original device.

The higher availability that results from this failover capability is not all that clustered architectures potentially have to offer, however. Relying on several devices rather than just one presents an opportunity to enhance system responsiveness and scalability as well.

Responsiveness comes chiefly from an ability to balance loads across the cluster, distributing storage, and reassigning, resizing or otherwise managing LUNs according to whatever set of policies have been put into play to manage the system. Scalability results from the cluster’s ability to add new nodes, or to add additional storage capacity to existing nodes. Ideally, this can be done with minimal impact to system performance – and hence, with minimal impact on the business processes the cluster supports.

The downside of clustering has always been the twofold challenge of designing an effective cluster, and then of providing an efficient method of managing it. Designing clustered architectures is not a job for the faint of heart. Microsoft has been trying to do it for years now with no notable success, and never progressed beyond clustering a few nodes. 

XIOtech claims to have solved the limited node issue for storage.  If its automated provisioning and intelligent storage management is also as good as it claims to be, it may have solved much of the management issue as well.

It will be interesting to see how all this plays out.  If the idea of seamlessly scaling your storage systems whets your appetite, check out XIOtech’s Web site (http://www.xiotech.com).