Virtualization Makes Illusion Real.

It's aggregating a lot of fans

It's remarkable what you can pick up during lunch. At the recent Storage Networking World, my publisher and I had lunch with Dick Blaschke, executive vice president at XIOtech Corp. The discussion turned to storage virtualization. This came as no surprise, since storage virtualization ranks with storage networking as highly-discussed, and more than a little controversial. The rise of large, complex, heterogeneous environments gives rise to the need to simplify management of storage, especially since storage requirements are growing at a rate of about 100% per year.

Blaschke drew a distinction during our discussion between storage virtualization and what he called "aggregation." He defined aggregation as the ability to centrally manage and share storage among heterogeneous hosts. Frequently, he suggested, people consider aggregation as true virtualization, rather than an element of virtualization. To be true virtualization, aggregation is only one element; the other is the ability to stripe across and utilize all available space in a centralized storage pool.

I asked for an example of aggregation. Blaschke pointed to the SAN appliance, which aggregates the storage that is on the SAN-regardless of what type of storage devices physically house the data--and gives the user a "virtualized view" of that storage. Because SAN appliances do not take virtualization a step farther by striping data across all available spindles, however, they suffer from many of the same limitations as other storage devices (direct attached, RAID). For example, the user must still be keenly aware of the physical characteristics of the disk drives within the system to effectively utilize and manage the storage pool. Also, because RAID levels are assigned in the individual storage devices seen in the "virtual view," the user cannot configure or migrate RAID levels from the SAN appliance box.

Still, what sets SAN appliances apart from traditional storage, is that the user can carve out space and perform LUN masking and LUN mapping on the aggregated storage from the SAN appliance box.

In both aggregation and virtualization, Blaschke said, a "virtual pool" is created that can be accessed and managed from a central location. The benefits of this come from data centralization and the ability to work effectively with a wide variety of servers and operating systems. Blaschke suggested that some products touted as virtualization products are just aggregation products.

This troubled me some, because selection of virtualization products is going to make a colossal difference in the way data is handled. The cold, hard, hype-free fact is that the end user...the employer of the integrator and VAR...does not have the staff to manage the crippling growth of storage in the IT installation. The lack of staff plus meteoric growth in storage screams for management. Lunch with Dick brought the issue into focus.

It's not as if virtualizing was something new. The concept of virtual disks runs back to the '70s. But a new variety of products has emerged, claiming to solve the problem for good and all. Let's try on a definition for size.

I spoke with Abbot Schindler at Compaq about this issue. Abbot identifies virtualization as imposing a level of abstraction between the operating system and physical storage. The level of abstraction permits a user to hand more storage at the same time. The value of virtualization, Abbot tells me, is in what the virtualization enables. Abbot's emphasis, though was identifying virtualization as instituting attribute-based storage management rather than merely physical management. Attribute-based management ties to the application rather than the actual physical device. The catch, as pointed out in a presentation from KOM Networks, is that applications aren't designed themselves for storage management issues.

I also connected with Claudia at Vicom in this most current examination of virtualization. She indicated: "Storage virtualization has become an essential service to manage heterogeneous SANs and enable true storage sharing and pooling. High data availability, efficient storage utilization, secure storage sharing, and non-disruptive backup are issues that storage virtualization services address through functionality such as mirroring, point-in-time copy, snapshot, virtual drive, drive sparing, zoning, and support for redundancy and multipathing. In addition, storage virtualization must also deal with the growing administrative burden of managing storage resources that continue to double every year. This is done through centralized administration and monitoring, a scalable architecture that supports growth in both storage devices and participating hosts, as well as accommodating multiple storage protocols and heterogeneous operating systems and storage vendors. The ultimate goal of storage virtualization is to reduce both the cost of acquiring storage resource as well as the cost of administration."

The definition makes sense to me, except for the conclusion that cost containment is the ultimate goal of storage virtualization. It seems to me that cost containment must go hand in hand with a measurable improvement in storage manageability. But, otherwise the Vicom definition works on most levels. Additionally, it identifies pooling as only one element of virtualization.

There are other attributes of virtualization to consider. Consider such activities as copy, swap, mirror data and perform LUN masking, LUN mapping, and clustering. If these functions are not automated, there seems to be no reason for virtualization... storage management will be as difficult as ever. Cost of physical storage is controlled, but the management of blocks or files (depending where the virtualization is located) is not addressed. The hiding of complexity...the creation of that level of abstraction is the issue.

A good argument could be made that the distinction is semantic only, so long as the basic definition of storage management simplification is accomplished. Whether it is abstraction, emulation, aggregation, host-based, server-based, static, or dynamic management might not matter.

What does matter is that intelligent selection of virtualization tools--be they hardware, software, or both--becomes practical. The potential buyer needs to ask about management of multi-vendor physical storage devices, limitations on I/O bandwidth, processor bandwidth limits, accommodation of a variety of server platforms and operating systems, capacities supported, and an assessment of what functions are operated within the virtualization engine.

Yes, it's important to tell virtualization from aggregation. But it's also important to decide whether aggregation will do the job alone, or whether more is necessary to accommodate applications. Those applications are still king...virtualization is just another tool in the struggle to keep those applications running effectively in the enterprise.