Chapter 7. Virtualization: Still a Dirty Word?

As discussed in the previous chapter, selecting the right disk and interface to meet the access and usage characteristics of applications is a basic element of intelligent storage architecture design. The differences in disk and interface performance ”and price ”also guide their inclusion in disk array platforms.

Arrays may vary in complexity from simple JBODs (just a bunch of disks), to more complex RAIDs (redundant arrays of independent disks) that afford a certain degree of internal protection against disk faults, to intelligent, multiported repositories that are increasingly acting as peer computers in a peer-to-peer I/O processing topology. The trend in the industry appears to be the delivery of arrays with the capability to scale their storage capacity dynamically through the use of an internal switch fabric. An increasing number are using Fibre Channel switches embedded in silicon chips to capitalize on the device connectivity capabilities of Fibre Channel fabrics inside the array, while relying on controller (or "head") intelligence to provide the operational management and control for the configuration.

Given the impending limits to magnetic disk areal density growth imposed by the superparamagnetic effect, the argument could be made that more capacious arrays, and ultimately networks of arrays, will be required to continue the pattern of increasing disk capacity at decreasing cost per megabyte that has long served as an engine of growth in modern information technology. To continue to scale disk beyond the areal density limit, multiple physical disk drives must be aggregated and presented as a single volume to operating systems and applications. A disk volume is an abstraction or simplification referring to a collection of blocks of storage. It is a virtual disk.