EMC, Isilon And The Rise Of NAS
I was a bit shocked this morning to get the news that EMC was buying scale out NAS vendor Isilon from the New York Post. The rumors had been flying that EMC was interested in Isilon for weeks but the Post, I wouldn't even wrap fish in the Post. That aside this acquisition bodes well for scale out NAS and in general for the pendulum to swing from block to file protocols for mainstream applications like VMware hosting.
November 15, 2010
Rumors had been flying for weeks that EMC was interested in Isilon, and this morning the news came down that EMC will buy the company. The acquisition bodes well for scale-out NAS. It also means an opportunity for the pendulum to swing from block to file protocols for mainstream applications such as VMware hosting.
High-performance NAS and scale-out systems like those from HP/IBRIX, Isilon and Panasas are among the highest performers and have been the predominant storage systems for applications in industries like media (especially CGI rendering), oil and gas geological analysis, and HPC (High performance computing) where a large number of compute nodes access a set of what are typically large files. Scale-out vendors compete with clustered NAS solutions from NetApp, IBM and the like. By contrast, commercial environments have been dominated by transactional applications, where a small number of database servers access a block storage system. NAS in the commercial space has been relegated to engineering departments and users storing Office files.
Over the past few years, the high-performance NAS market has become attractive to the big IT vendors. HP bought Polyserve, only to discover that a Windows-based product wasn't that attractive to the customers HP was hoping to serve, so it bought IBRIX as well. IBM built SONAS, and Dell bought Exanet, an acquisition that has yet to yield shipping product.
From where I sit, a big driver of these acquisitions is the use of IP storage protocols, especially NFS, for VMware hosting. NetApp has succeeded in getting the message across that it's easier to manage one large NFS file system with 100 VMDK files in it than 20 LUNs with 5 VMs each. Add in that Virtual Machine File System (VMFS) datastores are limited to 2TB, and a 200TB NFS share sounds better and better. I've heard reports from generally reliable sources that despite the talk about FCoE, many if not most of Cisco's UCS server installations use NFS for storage.
When you consider that a good scale-out architecture, like Isilon's, can scale performance, capacity or both as needed and the economics become really attractive. Overall, while some may think EMC overspent, the company has proven that the EMC stamp of approval, and legion of strong-arm sales guys in Armani suits and Gucci loafers, can take a startup's products and move it into corporate America.
About the Author
You May Also Like