Commentary: Storage virtualization is still in its infancy compared to its rich, mature potential. The goal of a ubiquitous, vendor-neutral storage utility will take more years of evolution of both the storage networking infrastructure and virtualization technology.
While doing research for a current book project on storage virtualization, I found it useful to keep a large bottle of aspirin handy for the headaches that were bound to ensue. Storage virtualization is the elephant that vendors and customers blinded by today's marketing hype lay hands on in total darkness and walk away with completely different impressions about what they encountered.
|“Separating what storage virtualization is from what it does is a useful exercise for understanding the common denominators of the fractional approaches vendors have taken.”
Attempts to make the technology understandable are sometimes helpful, but usually do not lift the veil of confusion from the subject. For example, in one explanation I saw, the mystery is reduced to the mundane by the declaration that virtualization is all around us. Even driving a car is a virtual experience since we don't have to get out and move the wheels every time we want to turn. Storage virtualization is, after all, just one variation of a very familiar phenomenon.
In another explanations, boxes representing storage, servers and a SAN are connected by lines of varying thickness, or by overlapping circles or explosive triangles. Arrows point to where the virtualizing business occurs, and sidebars explain the benefits. Despite these well-intended explanations, however, storage virtualization remains a complex technology whose intricate inner workings, ironically, are meant to make the life of the storage administrator simpler.
A Logical Look at Virtualization
The current market confusion over storage virtualization stems in part from the diversity of strategies for hiding the complexities of physical storage and presenting a streamlined logical view of storage assets. Storage virtualization creates an abstraction layer between physical and logical storage. That abstraction may occur in hosts, in storage arrays or within the SAN. It may also be performed in-band (a.k.a. symmetrical) with control and data on the same path, or out-of-band (a.k.a. asymmetrical) with control and data taking separate paths. In each case, however, there are no standard requirements for how storage virtualization must be implemented.
Unlike SCSI, Fibre Channel or iSCSI technologies, there are no objective standards that define how storage virtualization will perform. The closest the technology has come to standardization has been through the work of the Storage Networking Industry Association (SNIA). The SNIA's Technical Council and virtualization workgroup have created a taxonomy to explain basic concepts and associations. SNIA tutorials on virtualization have been presented at Storage Networking World conferences and are available online. Despite these noble efforts, the diversity of offerings and lack of standardization conspire to make storage virtualization a fertile ground for migraines for anyone who ventures there.
Market confusion over storage virtualization is also compounded by the intermixing of storage virtualization methods with the services that virtualization facilitate. Most virtualization vendor literature, for example, focuses on auxiliary services such as point-in-time copy (snapshots), data replication, utilization of storage capacity, support for heterogeneous storage arrays and so on. These storage services are enhanced by, but not dependent on, storage virtualization and so can be delivered by other means. Separating what storage virtualization is from what it does is a useful exercise for understanding the common denominators of the fractional approaches vendors have taken.
Continued on Page 2: Multiple Intelligences
Continued From Page 1
Storage virtualization intelligence is virtually all over the place. Host-based virtualization software provides independence from the SAN interconnect and storage assets, but must be administered on a per-server basis. Storage-based virtualization can often leverage other vendor-specific features/functionality, but typically lacks interoperability for heterogeneous storage environments.
SAN-based virtualization in the form of appliances or switch-resident intelligence offers centralization of virtualizing functions, but needs redundancy to provide high availability in the event of box failures. The market is still sorting out the strengths and weaknesses of each approach and there is as yet no overwhelming customer preference for a single solution.
In its current usage, the term storage virtualization applies more to systems than to discrete physical components. RAID, for example, is an atomic form of storage virtualization in that it presents a complex of physical drives as a single logical entity. RAID masks the complexity of individual drives and their geometries on the backend of storage systems while enhancing performance and data recoverability.
|“With so much brainpower somewhere in the SAN, it becomes feasible for applications themselves to dictate what their storage requirements are via APIs to the virtualization intelligence. ”
RAID is not a new technology, of course, and does not stir the imagination or whet appetite of customers the way that storage virtualization offerings have. System-level storage virtualization extends the abstraction layer over multiple storage arrays, hiding the complexity not just of individual physical drives, but of entire physical storage subsystems. The capability of intelligent virtualization agents to simplify management of storage systems amplifies the potential benefits of logical abstraction and provides a foundation on which higher levels of intelligence can be placed.
Storage as a Matter of Policy
The capability to treat multiple storage arrays as a single storage pool, for example, enables automation of repetitive and soul-numbing tasks such as LUN management. Automation of storage resources, in turn, accommodates a higher layer of policy selection and enforcement. Policy-based storage virtualization can govern use of the storage infrastructure to ensure proper allocation of different classes of storage. Priority applications get priority class of storage, secondary applications get less expensive classes of storage.
Policy-based intelligence is also a prerequisite for the next layer of application-aware storage virtualization. By monitoring data types, for example, storage virtualization can fine-tune data placement to more accurately meet the requirements of specific applications. Video data gets written to the outer tracks of higher performance disks; online transaction processing receives best-in-class snapshot care.
Lastly, with so much brainpower somewhere in the SAN, it becomes feasible for applications themselves to dictate what their storage requirements are via APIs to the virtualization intelligence. The parfait of functionality that spans from the abstraction layer, to automation, policy-based management, application-aware and application-responsive virtualization pushes storage networking towards its declared goal of becoming a ubiquitous and easily managed technology.
Virtually Potential Calls for Real Patience
Of course, none of this is rich and comprehensive functionality is available today. The ultimate promise of storage virtualization is still hovering somewhere in its own virtual reality. Current storage virtualization products achieve sundry bits and pieces of the ideal solution, but the technology is in its infancy compared to its rich, mature potential.
The goal of a ubiquitous, vendor-neutral storage utility as described by Compaq's Enterprise Network Storage Architecture (ENSA) and other industry literature will take more years of evolution of both the storage networking infrastructure and virtualization technology to achieve. The current state of the art, however, does have productive applications, as evidenced by the continued adoption of storage virtualization products for specific services such as point-in-time data copy. The confusion surrounding storage virtualization reflects its dynamism as an emerging technology, and while customers may have difficulty selecting the right solution for their specific requirements, due diligence and aspirin are highly recommended.
Director, Solutions and Technologies, McDATA Corporation
Author: Designing Storage Area Networks, Second Edition (2003) (available at Amazon.com), IP SANs (2002) (also available at Amazon.com).
See All Articles by Columnist Tom Clark