The virtualised information store

The virtualised information store

Enterprise storage is not exactly known for inter-operability. However, if storage virtualisation lives up to its promise, we ought to see storage products from multiple vendors communicating seamlessly with each other. by Anil Patrick R.

The strategy followed by most enterprises when it comes to storage purchases is to stick to the ‘one-vendor-offers-all’ approach—especially when it comes to high-end SAN boxes. This has been mainly due to the incompatibility of storage products from different vendors. In the past, such lack of inter-operability rendered helpless many a brave soul who dared experiment with a mix-and-match multi-vendor approach to storage. This scenario was endurable as long as enterprise data grew at a relatively slow pace. But as mega data churners such as enterprise wide applications began contributing to the information store, storage management across varied storage devices became a nightmare. It therefore became imperative that a single point of control to manage the entire organisation’s storage infrastructure be put in place.

With enterprises focussing on cost, the need to derive optimum value and higher application availability from storage has become top priority. It is difficult to stick to one vendor’s products due to the functionality demanded by applications.

ONE OF FIVE KEY ELEMENTS

The concept of storage virtualisation starts to assume greater importance at this point of time. Virtualising storage is not a new concept. IBM has been working on storage virtualisation since the 1970s. Even the traditional RAID approaches used in DAS are a form of storage virtualisation.

“Requirements like higher application availability, increased utilisation and increased productivity with a single point of control have been the main drivers for virtualisation,” says Subram Natarajan, Senior Solutions Architect, IBM South Asia.

As we shall see, the technology still has a long way to go. Storage virtualisation technology is currently restricted to the SAN environment. However, if the technology matures it has great potential. “There’s still a little hype, but virtualisation is one of the five key elements that will become a part of any comprehensive enterprise storage network. The five elements are storage resource management, storage network management, policy management, data management, and virtualisation,” says Sanjay Kharade, Principal Consultant, Cisco Systems, India & SAARC.

BEHIND THE SCENES

The basic premise behind storage virtualisation is to have a consolidated storage pool consisting of varied (heterogeneous) storage devices. The pool can be managed from a central console. It can be accessed as a single logical volume or distinct volumes as required. “Storage virtualisation can be used where an organisation has its data scattered over a number of devices. They can combine data and manage it centrally,” says Arindam Bose, GM, IT, LG Electronics India.

This is done by introducing a virtualisation layer that hides the complexity of heterogeneous storage. Known as block-level virtualisation in SNIA (Storage Networking Industry Association) parlance, this concept enables heterogeneous storage to appear as a single logical volume to application(s). It is possible to change volume or LUN (Logical Unit Number) size on the fly with storage virtualisation.

“Storage virtualisation is needed if you have a complex heterogeneous environment. It can be done at the server, switch and storage levels,” says Rahul Singh, Marketing Manager, StorageWorks Division, HP India.

THE ADVANTAGES

This single logical volume approach helps write applications with fewer inter-operability issues due to its device-independent nature. It helps reduce downtime since the application isn’t brought to its knees the minute a storage device bites the dust.

Higher utilisation, compatibility with multiple operating systems, and the ability to use storage from multiple vendors are all advantages conferred by adopting this approach. Storage virtualisation helps storage administrators perform backup, archiving and recovery easily, and in less time, by disguising the actual complexity of the SAN. Some of the other advantages include easier data migration, automatic capacity expansion, online disk-based recovery, and improved database performance.

This approach has several advantages on the operational side as well. For example, consider an organisation which has dedicated servers for activities such as reporting, accounting and Web services. These are attached to multiple storage arrays among which many might have spare space while others may be highly utilised. This creates a scenario where the total storage array capacity is utilised unevenly. Combining these devices is intended to help organisations create higher amounts of storage resources by optimally using existing resources. This also helps servers running different operating systems to store information on the same SAN.

If implemented properly, storage virtualisation can result in cost savings and higher availability. It can also help in mirroring requirements for DR purposes. “Pooling storage using storage virtualisation can help storage administrators improve the abysmal 40 to 50 percent capacity utilisation rates that typify most IT shops. That’s especially important for shops mirroring storage to a secondary site,” says Kharade.

THE FLAVOURS OF VIRTUALISATION

In a heterogeneous storage environment, current storage complexities can be solved with the virtualisation approach only with a thorough understanding of what it can do in a production environment Munish Mittal VP, IT HDFC Bank

Virtualisation mechanisms available today use dedicated hardware appliances or virtualisation software. The latter usually resides on fault-tolerant server hardware.

Host-based storage virtualisation is waning in popularity with virtualisation slowly moving into the fabric layer. “Virtualising the entire SAN is the latest storage virtualisation trend. SNIA refers to this as block level virtualisation; it helps heterogeneous storage devices appear as a single logical pool to applications,” says Ajaz Munsiff, Director, Virtualisation Products, EMC Asia Pacific.

In this context, the latest storage virtualisation buzzwords on the fabric front are inband (symmetric) and out of band (asymmetric) virtualisation approaches. Both use dedicated hardware to virtualise storage and have inherent pros and cons.

The inband approach is the older of the two technologies, and basically consists of a storage virtualisation device (appliance or controller) situated between the SAN switch and the disk arrays. The SAN traffic is routed through the device (usually a dedicated server) which provides the virtualisation layer. This approach is followed by vendors like IBM and HDS.

By contrast, out of band-based storage virtualisation relies on the intelligent features provided by layer 3 FC switches from Brocade, McData and Cisco. “Virtualisation is a by-product of the intelligence built into SAN switching devices. With SANs moving towards intelligent networks, virtualisation comes by default, thereby protecting the investment and reducing the cost of ownership of SANs. Intelligence includes routing and software services support on each FC port,” says Shuja Mirza, Technical Consultant, India, Brocade Communications Systems.

One of the biggest problems cited against inband-based virtualisation by out of band players is their potential for failure and slower I/O rates. Since the SAN data traffic passes through the device, there is a probability of having a single point of failure.

However, the fact remains that inband virtualisation is more widely accepted now since the approach has been around longer than out of band. In addition, virtualisation solutions for SMBs are provided only by inband players.

YET TO MATURE

As of now, storage virtualisation has yet to step outside the confines of a SAN. The term ‘storage virtualisation’ will become a reality only when the concept manages to integrate all storage devices used in an enterprise.

According to Munish Mittal, VP, IT, HDFC Bank, there is scope for more agreement among storage management vendors on the level at which virtualisation can be implemented. “It may probably work best when done at a level where data really resides. The technology requires some amount of a common framework before it becomes mature enough. This is why storage virtualisation is still at a nascent stage from a practical point of view,” opines Mittal.

The lack of standards is the biggest setback plaguing storage virtualisation today, although there is work going on for an SMI-S (Storage Management Initiative Specification). Initiated by SNIA, the standard is not attracting sufficient participation from storage vendors. Still in the process of evolution, the standard is an attempt to dictate cross-platform inter-operability among storage products. “SMI-S has yet to be widely accepted. What is happening is that vendors are choosing components rather than waiting for bodies to provide complete standards,” says Natarajan.

As of now, there are few takers for storage virtualisation in India. Indian organisations are still wary and are waiting for the technology to mature. “While companies are interested, nobody has yet implemented a SAN virtualisation solution as most vendors lack products and offer but a virtualisation roadmap,” notes Kharade.

Most Indian storage virtualisation projects are in pilot phase as of now. The reason behind this is that storage virtualisation still has to prove itself in production environments. “In a heterogeneous storage environment involving large amounts of data migration or data replication across storage sub-systems, current storage complexities can be solved with the virtualisation approach only with a thorough understanding of what it can do in a production environment,” observes Mittal.

Enterprises prefer to wait and watch for the technology to mature before they take the plunge. “We are still early even for large enterprise deployments (in the range of 50 to 100 TB). We will see more deployments by next year and by 2007 it will be mainstream. By 2008, storage virtualisation will be a done thing,” forecasts Munsiff.

As of now, vendors are suggesting the use of NAS headers to virtualise NAS devices. Till the technology matures to the point where it becomes possible to virtualise more storage devices than just the SAN, and standards are in place, the Indian enterprise is likely to be fighting shy of storage virtualisation.

Even if complete inter-operability does not become possible, storage virtualisation is maturing fast— and for the better—in that direction. Let us give it two years.

anilpatrick@networkmagazineindia.com