The fundamental idea behind storage virtualisation is to abstract, emulate and facilitate aggregation across different capabilities.

Today, several techniques are employed to virtualise different storage functions within the model. These include physical storage (layer I devices), RAID groups, logical unit numbers (LUNs), storage zones, LUN subdivision (a.k.a. “carving”), LUN masking and mapping, host bus adapters, logical volumes and volume management, file systems and database objects (such as a table space, rows, and columns).

Storage virtualisation as a technology is only useful if it offers better solutions to problems. The most severe problem that virtualisation helps solve today is the overall management of storage. Today, storage infrastructure represents one of the most heterogeneous environments found in modern IT departments, with a multitude of different systems at all levels in the stack—file systems, operating systems, servers, storage systems, management consoles, management software, etc. This complexity has become a hindrance to achieving business goals such as 100% uptime.

Problems related to data access impact several areas that can be improved by virtualisation are highlighted in the following paragraphs.

Single points of failure within the SAN, whether in a single array, a switch or some other component of the path from application to storage, are always a problem and without virtualisation, eliminating them can be prohibitively expensive.

One of the key metrics for judging quality of service (QoS) in the SAN is performance. Today, maintaining deterministic storage performance for applications is a very complex task. Virtualisation offers the opportunity to improve performance, but more importantly it offers the chance to manage performance in real time to maintain guaranteed and measurable QoS.

The data stored in the SAN is critical to the day-to-day operation of a company. Losing access to it, even for a couple of days, can be disastrous. Virtualisation is a key enabling technology in making more affordable options for disaster recovery and data archiving available to customers that can’t afford today’s high-end solutions.

In addition, the ever-increasing demand for “raw” storage (i.e., level I devices that supply capacity) is partly driven by poor utilisation. Poor utilisation inevitably leads to unnecessary expenditures in both hardware and management costs.

Industry surveys indicate that capacity utilisation in open systems environments, is at 30–50% for disk devices and 20–40% for tape. This means that companies are, on average, buying 2–3 gigabytes (GB) of disk capacity for every gigabyte of stored data. Clearly, these rates are not viable for an ongoing business.

Majority of Enterprises in India today face the challenge of poor utilisation of storage capacity. The root cause of poor utilisation is the way in which storage is allocated and bound to hosts. As the process is complex and prone to error, storage administrators tend to allocate large chunks of capacity simply to minimise the number of times they have to repeat the procedure during the operational life of the host. Storage virtualisation offers new ways to address this problem through automation, just-in-time provisioning, and tiered storage architectures in which different quality levels of storage (i.e., low-cost storage for non critical data) are dynamically allocated to meet specific requirements using predefined and automated policies.

Storage-based virtualisation is typically not dependent on a specific type of host, allowing the array to support heterogeneous hosts without worrying about variance of host operating systems or applications. Also, storage-based RAID systems deliver optimum performance in relation to their hardware because features like caching can be tuned to the specific hardware. The downside to this approach is that the storage virtualisation functions are typically confined to a single array; for example, the source volume used for a snapshot and the snapshot itself are maintained on the same array, making the snapshot useless in case of hardware failure. In some cases, virtualisation functions extend across multiple arrays, or a cluster of arrays or controllers; however, these solutions are typically restricted to a single-vendor implementation.

With the proliferation of remote and mobile workers with their data scattered across multiple sites, it becomes increasingly important for virtualised environments to span the scope of a company's network; be it a campus, a state or a distant branch office. Virtualisation has no reason to be confined to a box or a room. In fact, it should supply affordable redundant copies of vital information in separate locations for consistent access to the same data by any number of authorised users. Virtualisation should be independent of the LANs, WANs, and SANs that link storage farms together. That is the whole idea behind divorcing the applications from the storage hardware.

Storage Virtualisation Considerations

If you are not already leveraging or planning on deploying some form of virtualisation technology to address your data and storage management challenges, the following questions will help you plan and be better equipped to sort out the various options.

Keep in mind the issues or challenges you are looking to address with virtualisation technology. For example, are you looking at volume, LUN or file system pooling and aggregation to improve capacity utilisation, enhance interoperability, reduce management complexity, reduce costs or eliminate real or perceived vendor lock-in?

Ask yourself if you really need or simply want virtualisation technology. For example, you may want to have virtualisation to create a seamless transparent pool of unified storage to reduce costs and eliminate vendor lock-in. Or you may need virtualisation to enable transparent data migration and movement for tiered storage or to eliminate downtime caused by technology upgrades or disruptions from capacity movement and allocation.

How will virtualisation technology fit into your existing environment? Look at what procedural and configuration changes will be needed, along with training and education requirements. What other technologies will be required on host servers, including applications, agents, managers, drivers, shims or other forms of software?

Understand how a candidate virtualisation solution or technology will scale in terms of number of storage devices and servers attached, overall capacity, performance (IOPS, bandwidth, latency) capabilities, ease of use and availability.

Virtualisation technologies and solutions can have great benefits for almost any IT organisation regardless of size. Specific types of virtualisation solutions and technologies may have more relevance for specialty environments. When it comes to virtualisation technology, start simple and evolve, understanding your scaling issues, and keeping in mind what the specific virtualisation technology is designed and intended to support. Realising that different solutions are targeted at various functions, and then aligning those capabilities to your needs, can be the key to a successful storage virtualisation technology deployment. Done right, virtualisation technology should be transparent and reduce your management activities and data center complexity.

Way Forward

The future of Storage Virtualisation lies in unified management with integration of discovery, reporting, storage virtualisation, and storage automation. In a heterogeneous world, a method for intelligent storage provisioning through active management at all levels, from the application to the physical storage devices, is only feasible in the long term with virtualisation working together with open standards such as SNIA SMI-S. “Storage as a utility”—heralded years ago when the first SANs appeared—can soon become common.

The author is CTO, India Of Symantec Corporation