Wikibon Blog

"You must admit there is a fit..."

Storwize CEO, Ed Walsh stopped by the Wikibon offices the other day with his new VP of technology strategy Steve Kenniston packing a new pitch. Wikibon got an early look and we were definitely impressed. I first learned about Storwize in late 2008  and then mid last year we had Burzin Engineer of Shopzilla on a Peer Incite along with several other practitioners, speaking about in-line data compression for primary storage.

What I didn’t grasp until just recently was the potential that Storwize has as a company. Frankly I was surprised at how well Storwize had been keeping its secrets. I think Walsh and company have plans to change all that. For those who haven’t been following storage closely over the past few years, Walsh was the CEO of Avamar and then Virtual Iron. He has a track record of sniffing out firms with valuable IP in white hot markets,  building organizational structures, growing a business and then selling it to deep-pocketed firms like EMC and Oracle. But let’s not get ahead of ourselves– there’s still plenty of work to be done here.

But there’s also a much better foundation to build on than I had realized. To wit:

• Storwize has over 100 customers in the Global 5000.
• In addition to Shopzilla, there’s Chevron, Polycom, Mazda and Allianz and a bunch of other household names that aren’t public.
• Storwize has a bagful of patents ~35 either approved or pending with some pretty interesting secret sauce, solving a challenging problem; namely compressing primary storage without sacrificing performance.

The company has an architecture called RACE– Random Access Compression Engine which provides real time (in-line), lossless data compression explicitly designed for primary storage applications. Customers on the Peer Incite were seeing compression rates of between 50-90% and Storwize shared with me sampling statistics from the field that corroborate these data points. The other thing they shared with me is proof points that IBM has developed validating the fact that Storwize compression doesn’t degrade performance. When you dig into the IBM study you can see that Storwize actually increases performance. The marketing people from Storwize are reluctant to make this claim but the handful of Storwize customers I’ve spoken with emphatically says this is the case.

The reason is pretty interesting actually. Storwize compressess data on an initial write and as a result, the cache in the disk array controller is effectively enlarged, meaning more data will be serviced out of cache. This increases cache hits and minimizes cache de-staging to backend disk; thereby increasing performance.

What makes this even more compelling is we’re seeing a steady pattern in the industry as it pertains to data compression. The technology is consistently moving further up the storage hierarchy. It started with tape compression more than a decade ago with companies like StorageTek and IBM. Then companies like Riverbed popularized WAN optimization earlier last decade. Next data de-duplication from companies like Data Domain and Avamar exploded onto the scene and now companies like Storwize and Ocarina are using the speed of modern processors to power much more sophisticated compression algorithms than the computing industry has been able to process up to this point.

While there are clear differences in the types of compression technologies and use cases that are being brought forth for primary storage, it seems clear that the trend to blindly keep storing non-compressed data must end. Storwize for its part has much to do building a brand, improving its service capability at scale, increasing account penetration and ideally securing some OEM partners for its technology. Ultimately, what will make Storwize attractive is an OEM partner is the ability to embed the Storwize IP into its stack. If it can figure out how to do this it will have a leg up in the race to compress.

I need to dig into this more to understand the applications for various types of compression. It’s clear to me that for example while Storwize and Ocarina are solving different problems, they’re both on the curve to the next generation of  data reduction technologies. This is important because in this era of do more with less; when it comes to data compression for primary storage– you must admit, there is a fit.