Storage scalability in a virtualized environment is quickly becoming a concern for some data centers. Unlike the very predictable single server world which had a single application, single NIC card and single host bus adapter, the virtualized host can have dozens of virtual machines and multiple network interface cards. This leads to a very unpredictable and random workload that can push storage controllers to their limits.As we discussed in our latest video, "Dealing with the Storage I/O Blender" the server and network interface side of storage I/O optimization can be handled by Host Bus Adapter QoS. On the storage side all this I/O still gets to the storage controllers and can quickly saturate the traditional dual controller storage architecture.

To make matters worse, most "active-active" storage controllers are not what they appear. A LUN or group of disks is only assigned to one controller at a time. While each controller can be different LUNs (that's the active-active part) both controllers don't share the workload for a given LUN. The alternate controller can act as a standby for the other controller. It is a failover option not a performance option. This can lead to, especially in larger virtualized environments, performance problems. To solve the controller bottleneck some storage vendors have developed grid or cluster based storage systems.

A grid or cluster based storage system is typically made up of a series of nodes or controllers that can be added to the system to scale performance as the workload demands. As we discussed in our entry on Network Computing, "Tightly Coupled vs. Loosely Coupled Clusters", tightly coupled clusters tend to be more performance oriented but have more strict compatibility requirements of their members.

These grid-like architectures are available now for almost all categories of data centers. At the enterprise there are solutions from 3PAR, IBM and Isilon. In the mid-range of the market there are solutions available from Scale Computing, Pivot3 and HP's Lefthand Networks products.

Most of these solutions scale as you add capacity. Each node that is added to the system brings with it capacity, I/O performance and I/O bandwidth. It is important to understand how this additional storage and capacity becomes available. Is it automatically added to the storage cluster and then all the connected servers see the additional capacity and performance or do you have to manually assign that capacity? Either option may be fine for your environment, but understand what is involved in gaining access to these additional storage resources when you plug them in.

Grid or cluster based storage systems involve a little more work upfront to build the foundation of the cluster, although many vendors have taken the work out of that as well by pre-packaging the initial group of nodes. The payoff of extra upfront work can be a significant saver of capital dollars and operation dollars as the solution grows.

Track us on Twitter: http://twitter.com/storageswiss

Subscribe to our RSS feed.

George Crump is founder of Storage Switzerland, an analyst firm focused on the virtualization and storage marketplaces. It provides strategic consulting and analysis to storage users, suppliers, and integrators. An industry veteran of more than 25 years, Crump has held engineering and sales positions at various IT industry manufacturers and integrators. Prior to Storage Switzerland, he was CTO at one of the nation's largest integrators.