This week, it's the Flash Memory Summit and I'm sure I'll have an interesting dispatch from there once it wraps. The event comes at what feels like a pivotal time for the solid state industry, especially in the mid- and high-end business IT markets. This is simply because the basic value proposition of flash/SSD/solid state storage (pick your moniker) seem to have been established in the minds of enough buyers that the various solid state technologies are really beginning to get traction. What was once seen (oh my, doesn't three years ago or so, when the current flash era really began for data centers, seem like ancient history!?) as just a cool niche technology is now turning into a real market that's definitely heating up. This means that the debate will gradually shift away from the basic capabilities and foibles of the raw technology to the more business-meaningful attributes of high performance claims, implementation styles and, yes, even value for money.
A couple of recent examples: in terms of high performance, we had the news at the end of July that IBM had used Violin's drives to help it break its own GPFS record for file processing by an impressive 37 times (going from one billion files scanned in 3 hours to 10 billion files scanned in 43 minutes); just today we have the news from Texas Memory that it can deliver one million IOPS (random, 100% read, 512k blocks, single server) through its RamSan 70 450 GB PCIe card. Performance is one side of the solid state coin, the other is implementation. While the storage system vendors are focusing on using flash as cache and/or integrating it with automated tiering, there was also an interesting move last week from Fusion-io. It announced the acquisition of a small but nifty start up, IO Turbine, whose software will allow Fusion to begin to essentially share its solid state drives across virtual machines, helping drive consolidation and VM density.
The battle lines are being drawn. It seems that the first stage of the war (that solid state can help IO, whether as overall throughput or outright performance) is done and won ... now it's about what type of solid state you use and how you use it. Already there are vendors offering fixed-data solid state (where for example Fusion- io was), shared solid state, variants as cache (many of the systems vendors, such as EMC, HP, IBM, HDS, Dell and NetApp play here), and there are emerging software only vendors (such as VeloBit and FlashSoft) to optimize all the above. Beyond this--some now, some soon--we'll have hybrid or all-flash arrays (from the likes of Pure, Nimbus, SolidFire and Nimble) which will extend the "actually good value compared to high performance spinning disks" weapon.
Interesting times for sure. The cool tech is definitely becoming a heated market.
Read more of Mark's blog entries at The Business of Storage.
Browse by Content Type
Categories
Share