 |
Randy Kerns
Partner, the Evaluator Group
|
 Randy Kerns is a partner at the Evaluator Group and is responsible for storage area networks (SAN) and network-attached storage (NAS) analysis and education as well as company and product strategies. He has over twenty-eight years storage product development, including work for IBM, Fujitsu, Vice President of Engineering at the Array Technology subsidiary of Tandem Computers and Director of Engineering for Enterprise Disk at Storage Technology Corporation. |
|
|
Storage virtualization solutions that work at the fabric level have been shipping to customers from vendors who have been ahead of the curve for over three years now. In that period of time, we've had the opportunity to watch solutions mature, advanced features get added and the real value for customers emerge. Now, additional vendors are bringing their solutions to market, so it is worthwhile to look at the successes and look forward to the future for these types of solutions.
It is important to understand that storage virtualization itself isn't the end product being offered to customers. It is an enabler for other capabilities which have been promoted in various marketing-speak materials. The value in fabric-based solutions generally can be thought of as the ability to abstract storage to present a common management interface for storage management software, and the ability to add data movement functions at the fabric level for such functionality as remote copy and point-in-time copy regardless of the type of back-end storage and without using special storage device capabilities.
The common treatment of storage enabled by fabric-based virtualization provides many advantages: It makes the task of storage provisioning much easier with a very realistic opportunity for automation, it allows movement of data between storage systems for load balancing and to appropriate performing devices based on the data requirements, it centralizes the reporting and data collection as well as the basic management functions (rather than having to use a special piece of software or API for each device type), and it enables many of the additional functions that companies have talked about for quite a while.
The fabric-based functions of remote copy and point-in-time copy are going to prove very valuable to customers. PIT copy that has been integrated into storage systems for a number of years, starting with the StorageTek virtualized storage system (now called the SVA), has demonstrated great economic value in providing data availability for backup, testing and other environments. Doing it at the fabric level will allow different types (and classes) of storage systems to be used. For example, a clone copy could be made on a cheaper and lower performing system than the source system. Remote copy is becoming more of a requirement in businesses. Doing it in the fabric allows the remote copy to be done between heterogeneous devices and without the individual controls and charges for specific storage systems.
The successes seen so far with the more than 1,600 systems installed have been impressive, and customers have integrated them into their business operations. The primary sales have been through the distribution channel and targeted direct sales.
IBM has now entered the market with its own internally-developed product, the SAN Volume Controller and a packaged, virtualized SAN-in-a-box solution, the SAN Integration Server. IBM is also initially focusing sales through the channel using business partners. IBM is a big-name company that is trusted by most executive management so they should be able to expand the success of fabric-based virtualization. When they begin to drive direct sales into their high-profile accounts, the visibility will be even greater.
Other vendors are either entering the market or are poised to enter with differing solutions. Some plan on using volume manager software in an intelligent switch that cracks open frames to redirect access. The momentum and successes so far are with the specialized controllers (usually based on standard server platforms) that are in the data path that generally use caching to provide performance acceleration, which more than compensates for added latency.
The conclusion we have to draw is that the last three years have seen fabric-based solutions start slowly and grow in successes with very referenceable customers. Now, the big name companies such as IBM and HP are pushing solutions with others soon to follow. The abstraction of storage or virtualization at the fabric level will be a very viable storage solution that is "prime time."
For more information:
Q&A: EMC's Gahagan on taking a virtualization ride
Article: Despite its age, virtualization lauded as next best thing
Expert response: Why and how to virtualize inside a switch
