Storage virtualization: In-band or out-of-band

The best answer to the question may be, "It doesn't matter."


Why is it that whenever the topic of storage virtualization comes up, a debate over the merits of in-band versus out-of-band-or where virtualization is implemented-often follows? These debates have become so intense that end users could be led to believe that every storage strategy should pivot on this single issue.

Virtualization is an important technology that will enable users to finally realize the business benefits of network-based storage, but to dwell on whether the architecture is in-band or out-of-band (symmetric or asymmetric) completely misses the big picture. While some storage vendors and analysts will try to spin this tactical issue into a strategic one, how block- or file-level data is virtualized becomes meaningless in the grand scheme of data storage.

While vendors like to tout reduced total cost of ownership (TCO) and enhanced return on investment (ROI) through the use of storage networking, when it comes to drawing a correlation between these benefits and how a solution virtualizes data, the line is fuzzy at best. The tendency to tag every solution according to how it virtualizes relates only to its architectural approach, but not to the value that the particular solution provides. All of these architectures can deliver virtualized storage to application hosts, and most have the potential to lay a foundation for the next level of benefits beyond those of basic block-level virtualization.

The various virtualization approaches can be easily distinguished by labeling them in some arbitrary fashion, but this means little when you consider IT's ultimate storage goals. All approaches have inherent strengths and weaknesses, but they're rarely quantified in terms of how certain architectural attributes contribute to an enterprise's goals, such as business continuity, new line-of-business functions, policy-based asset management, and ROI.

Every type of storage system has some degree of in-band control, and every network and file system has some degree of out-of-band control. If you can envision a "storage dial tone," or the so-called "storage utility," then the underlying technology will in all likelihood be a hybrid of both. The storage dial tone/utility implies not only block-level virtualization but also a virtualized file system. As such, it makes more sense to consider how today's virtualization options will fit into the file-level scheme of things, rather than scrutinizing whether control flows over the data path or via a separate control path

A virtual file system requires that hosts are aware of it, meaning that software/ firmware must be installed on the hosts (out-of-band control). Does the file system really care if block-level storage is virtualized via in-band or out-of-band techniques? Does it care if the topology is appliance-, switch-, or router-based? Not really, as long as the interface works. Whether in-band or out-of-band, what is vital is how well today's storage area networks (SANs) can fold into the next storage paradigm.

Selecting a SAN solution today involves many factors, and different applications may fare better with certain types of configurations. What really matters is finding the combination of storage, network, and system management technology that provides optimum value to the user in his or her application environment.

Consider the strategies of some of the big platform and software vendors in the enterprise storage space. Although some of their offerings already encompass in-band and out-of-band SAN appliances, their marketing emphasis has begun to shift to enterprise-wide storage visions, now that definitive products are on the horizon.

For example, IBM's Storage Tank will deliver a true file-level virtualization scheme in the form of a SAN operating system-like file system. Compaq's VersaStor initiative complements IBM's Storage Tank by delivering block-level virtualization for a combined solution with application (block) and distributed (file) level virtualization granularity. Veritas is also going after the enterprise with this same type of capability. While block-level virtualization is currently available from more than a dozen companies, enterprise-level solutions are well down the road for most solutions providers.

The benefits of SANs are too compelling to wait for the rest of the virtualization story to unfold, but trying to decide what kind of virtualized block-level SAN to implement can involve a leap of faith when evaluating its ability to deliver the next level of storage benefits. Any enterprise aiming to implement one of these futuristic SAN file systems should determine the ROI of what's on the table today, with due weight given to how these SAN pieces fit into the long-term IT strategy. In light of everything that goes into a successful storage strategy, the issue of how and where a SAN solution virtualizes data should be relegated to a shrug and the response, "It doesn't matter."

Richard Lee is president, and Harriett Bennett is a senior analyst, at Data Storage Technologies, a strategic marketing services consulting firm in Ridgewood, NJ.

This article was originally published on June 01, 2001