Where should storage intelligence reside? The answer is everywhere. It all depends on the needs of the environment. For some customers managing data will reside on host servers. Other customers will continue to use their storage systems to perform these functions. An emerging approach is to use a storage-virtualization product. And of course, many customers will use all of the above.
Is storage virtualization an emerging approach? SAN virtualization has been around for years but there are relatively few implementations at this time. ESG estimates there about 3,000 storage-virtualization products installed in production environments. That is what I would consider 'emerging.'
Typically, the SAN-virtualization product sits in front of heterogeneous storage systems. The applications see the SAN-virtualization product as its disks and not as the storage systems any longer. For example, if you're using a storage system from Vendor ABC, the applications would see Vendor ABC volumes. However, after installing a SAN-virtualization product from Vendor XYZ, the application would then see Vendor XYZ volumes. This means that data has to be migrated and re-mapped to the SAN-virtualization product. This is what I call full virtualization for SAN-based storage.
Partial virtualization is another approach that can be extremely useful for performing local and remote mirroring functions. The storage-virtualization product does not alter or change the format of the data or present itself to the application as the storage system. Instead, it can take a volume stored on heterogeneous storage systems and copy that data to another heterogeneous storage system. This approach is nondisruptive and can reduce costs significantly.
There is also an out-of-band approach that requires customers to install agents on the host servers. The storage virtualization product is not in the data path but writes are sent to both the storage systems and the storage virtualization product. In this case, the storage virtualization product does not sit in front of the storage systems but to the side of them, if you will.
Today, appliances are the most popular approach to providing storage virtualization. These are typically Intel-based servers that run storage-virtualization software and have been hardened to provide a turn-key product that is reliable and secure.
Customers have been communicating to ESG for years that they want storage virtualization capabilities in a switch. The rationale has been that since the SAN switches are already in the data path, the switch is the right place to put storage virtualization. However, the software intelligence required for storage virtualization is not the core competency of storage switch vendors. With that in mind, the switch vendors' approach is to provide a platform for third-party vendors to integrate their software products into the switches. Storage software vendors can implement their software on a blade that is installed into the SAN switch. Another approach is to have the storage-virtualization product remain in an appliance or running a server and use the SAN switch as a data mover using fast-path technology.
The latest approach to SAN virtualization is a hybrid product that is both a storage system and a storage-virtualization platform. These hybrid products are storage systems that can be used just as storage systems. Additionally, they can be used to manage external heterogeneous storage systems. The advantage of these products is that they can provide two capabilities in one. Customers who need to acquire a new storage system can also use it for storage virtualization instead of buying discrete products for each function.
NAS virtualization is a rapidly emerging area that is moving into the spotlight. Customers often have dozens, hundreds and even thousands of NAS file systems that create complex environments for users to navigate around. There is also a burden on the IT group to manage all of these discrete file systems dealing with adds, moves and changes. Additionally, NAS-virtualization products can create a single logical view across all of the file systems across multiple NAS systems and is able to nondisruptively migrate data from one storage system to another. This becomes important for more efficient capacity utilization and improving performance.
NAS virtualization is extremely transparent. These products do not have their own file system but are an intelligent layer in front of the NAS-storage systems. Therefore, all of the intelligence and value that you get from your NAS products is protected. NAS virtualization for the most part improves upon but does not replace any of the intelligence provided by the NAS systems. However, some NAS-virtualization products do provide remote copy services supporting heterogeneous NAS systems.
Storage virtualization is still an emerging segment but the signs are all there that it is now being embraced. Practically all of the large storage system vendors now have some storage-virtualization product. Customers that have implemented these products have quantified the value and have realized enormous cost savings ranging from hundreds of thousands to even millions. There will be confusion in the market for some time as customers figure out which approach suites them best. Does it make sense to use an appliance or a switch or a hybrid product? Should I start with SAN or NAS virtualization? How much of my storage network do I virtualize? Do I even bother for now and just maintain status quo? One thing that is important to realize is that the customers that have implemented storage virtualization, after careful planning and in some cases trial and error, are reaping the rewards.
About the author: Tony Asaro is the senior analyst for Enterprise Strategy Group.