IBM has purchased the oldest and most respected high-end SSD builder in the country: Texas Memory Systems. Founder and CEO Holly Frost has led the company for 34 years. No word on price, but I expect it was not stratospheric.
Before the advent of flash memory, the SSD business seemed simple enough: stick a bunch of DRAM in a box with a battery backup; add the I/O channel du jour; and tout the performance – while ignoring the price. Many companies tried – and failed – but TMS soldiered on.
While SSD technology was important, the real key to success was building successful customer relationships. And that meant finding the critical applications of affluent customers and understanding how best to use the costly DRAM SSDs.
Flash SSDs
NAND flash changed the traditional SSD business by making SSDs much more affordable. TMS moved into flash SSDs, but lost out to nimbler competitors such as Fusion-io and Violin Memory.
Good technology but limited market penetration and a company too small – ≈200 employees – to grow downward from their traditional high-end niche. Employee morale began to drop and good people started leaving for companies where their years of SSD expertise had a better chance of reward.
Enter IBM.
The StorageMojo take
This is a good move for both companies. TMS gets the distribution to take their considerable expertise to a much wider market. IBM gets a proven product with excellent references to feed their large customer base.
It isn’t a billion dollar IPO, but it is a win for both and the larger storage market.
Courteous comments welcome, of course. I recently joined IBM’s Storage Community site – stay tuned for my 1st post – so maybe I’m now biased. OTOH, I can’t see how the TMS buy could be a Bad Thing. Perhaps a reader can enlighten me.
Couldn’t agree more, this is a great move for IBM. There is no downside.
I expect though that it will be TMS’ non-volatile DRAM based SSD technology that will be most important to IBM, and not the Flash-based stuff. TMS has some great IP around destaging DRAM to disk/flash in the event of a power failure, and the real heavyweights in financial and trading markets have standardized on DRAM SSD for years. They simply can’t tolerate the hideous write latencies of NAND flash.
For example, the DRAM based RamSan-400 is half the cost per IOPS of any Flash-based SSD solution, and with 1/10th the latency! (SPC-1).
As everyone who is marketing SSDs already knows, SSD value proposition is all about cost/IOPS and low response times, not about gross IOPS and cost/capacity. DRAM wins on both fronts, and it’s not a close race.
It’s like Jim Gray (an IBM alumnus himself) has said… “tape is dead, disk is tape, flash is disk, but DRAM locality is king”.