• Home
• Articles Archive
• Features List

• Media Pack
• Contact Us
• Subscribe

Virtualisation all wrapped up

From STORAGE Magazine Vol 6, Issue 9 - December 2006

Most of the business benefits of virtualisation can be gained with standard disk arrays or heterogeneous virtualisation appliances. now Users have begun to find 'real-life' situations that specifically apply to the latter, according to Robert E. Passmore and April Adams of world-renowned research organisation Gartner

Heterogeneous storage virtualisation systems have been deployed for a variety of purposes. The most frequent reason is to facilitate the migration of volumes from one array to another. This occurs most typically when transitioning from old, obsolete equipment to new disk arrays.

Previously, migration was mostly done with server-based software tools, either breaking mirrors or migrations, but virtualisation appliances enable the entire migration to occur from a single control point, without disrupting applications. This also takes fewer staff and is less prone to errors. IBM, with its SVC, has been particularly successful in exploiting this use case. EMC's Invista, which came much later to market than the SVC, has had functionality mostly limiting it to this case.
The second most popular use case at the upper half of the market is remote replication. On the low end, however, several vendors, such as DataCore, have gained market traction from providing lower-priced replication alternatives. Given that most standard arrays do remote replication only between like models, the promise of the virtualisation devices seems to be to allow remote replication to dissimilar models. We would point out, however, that, in most cases, the virtualisation appliance is doing the same thing as conventional disk arrays - that is, replicating to a like appliance at the second site - and that the presence of heterogeneous storage is again with inexpensive basic disk or more-expensive disk arrays being used as inexpensive basic storage behind the appliance.

The third most popular reason has been to get lower-cost disks into higher-performance environments. HDS (and to a lesser extent partners HP and Sun) has been particularly successful in implementing this use case, as it has enabled midrange Advanced Technology Attachment (ATA) disk systems to be integrated or virtualised into the high-end USP/NSC family of arrays. We note that the USP/NSC disk arrays do not currently support low-cost high-capacity disk drives internally, whereas competitors IBM (DS8000) and EMC (DMX3) do.

The fourth, and least common, use case has been immediate consolidation of diverse storage environments into a single storage area network (SAN) environment, using the virtualisation appliance as the primary array controller and source of value-added software moving forward.

Most of the users deploying this use case describe their previous environments as "out of control" from a storage perspective, in the sense that storage was bought previously by different parts of the organisation, from different vendors, and now needs to be brought quickly under control.

In this case, the virtualisation of heterogeneous disk arrays is an attractive, but temporary, transition move that will result in replacement of these disk arrays with basic, inexpensive storage of choice. In the meantime, users can get additional use from their legacy resource, and move quickly to a unified automation and management deployment. The result is that the appliance becomes the array controller and source of value-added software, and the user is typically locked-in to the supplying vendor.

That mobile feeling
The most important benefit provided by virtualisation is mobilisation of the virtualised resource under software control. This simplifies business operations and storage management, and reduces management effort by enabling issues associated with installing, configuring, monitoring and managing physical storage to be isolated to a small group of well-trained storage administrators. It also eliminates the need for other IT personnel (such as application managers and systems managers) to be aware of, or trained on, the problems of managing physical storage.

With the right software/management tools to take advantage of virtualisation, the business gains flexibility and agility in provisioning, maintaining availability, recovery and other critical storage processes. Administrators are no longer constrained by the physical limits of the disk arrays, and can build the quantity and size volumes that they require. Furthermore, virtualisation can offer dynamic volume change capability. Administrators don't have to guess what each application's requirements will be down the road and then either live with the results of their estimates or undergo time-consuming, disruptive and expensive reconfigurations.

With today's virtualisation capabilities, administrators can make transparent changes in volume size within an array and it is expected that, in the future, it will be possible to build volumes across the boundaries of arrays and vendors.

Quality of service encompasses the areas of availability, performance and cost to an application. Virtualisation is intended to provide the right mix of these elements by:

• Eliminating the stranded capacity problem associated with non-virtualised arrays (increasing use)
• Avoiding 'hot spindles' (disk drives experiencing intense input/output loads, because of frequent access requests or because the redundant array of independent disks - RAID - level assigned to the drive increases the drive's workload) and simplifying the process of redeploying capacity when the needs of an application change (improving performance)
• Increasing performance by striping and caching
• Providing easier migration to new subsystems
• Simplifying the addressing of storage volumes (increasing flexibility)
• Reducing downtime, because changes are made through storage management tools at the virtual level, rather than at the physical level (increasing availability)
• Reducing staffing and training costs, with more capacity managed on fewer storage devices (cutting cost)

Most of these benefits can also be associated with legacy disk array controllers and with heterogeneous virtualisation products.

Management tools
Virtualisation should present a simple object (such as a volume) upward in the stack and hide the physical complexity of networks, storage and other constructs. Management tools, however, should be designed with a different goal - to provide visibility for all layers, including the mapping implied by virtualisation. If the application fails, then the management tools would be able to report the location of the data and all the hardware in between.

Storage management software promises to provide capabilities such as path management, performance optimisation and provisioning, which automate the challenges of managing through all the layers. However, virtualisation software could block the visibility of the management tools.

Virtualisation is an important middle layer in most vendors' visions for the future of storage management. These visions promise applications that can automate policy throughout an enterprise environment. Single-box behaviour provided by virtualisation for the entire storage pool in an enterprise SAN is key to simplifying the development and operation of these applications.

Most of the heterogeneous virtual- isation appliances shipping have problems in this area, however. Some purport to support other vendors' disk arrays, but do not integrate the element managers into the appliances's software. Few, if any, are supported by comprehensive SAN management packages from other vendors. Gartner strongly recommends that enterprises evaluate the availability, compatibility and level of integration provided by virtualisation, and other 'network intelligence' software, in relation to other management tools.

As with any evolving technology, the benefits promised by virtualisation are subject to a great deal of vendor hype. Users should carefully sort out the claimed benefits of products to ensure that they are not being implemented redundantly (in, for instance, an appliance and a storage array) and that they don't create more management complexity to get the implied benefit.

THE RISKS AND THREATS
The challenges related to virtualisation are associated with the extra layers of software, and potentially with extra hardware, that threaten to increase both cost and complexity. The additional code and components can potentially reduce reliability. If the virtualisation is in the data path, for example, the appliance becomes both a single point of failure and a potential limit to performance. Another risk stems from the often disruptive installation and configuration of these devices.
Finally, when virtualisation is in the data path, hiding the physical complexity from the application, it can get in the way of manageability. ST