The only primary storage data reduction product we examined for this story that doesn't have a deduplication component, Storwize Inc.'s STN-6000 inline appliance focuses strictly on real-time compression and installs in front of network-attached storage (NAS) filers from vendors such as EMC Corp. and NetApp Inc. Storwize claims that the product provides compression ratios of 2:1 to 15:1, yet its major distinguishing characteristic isn't compression algorithms.
CEO Ed Walsh said that compression is typically sequential and slow on random reads and writes, and the system needs to decompress the whole file to do one I/O. With Storwize's technology, the system can operate on data in read mode and write mode, random access, without compressing and decompressing the entire file, he said.
"That might seem very simple, but that is really hard, and that's the difference between being performant real-time, random-access disk," Walsh claimed. "That's the essence of why we have no performance degradation, but we get the same compression ratios."
Lauren Whitehouse, a senior analyst at Enterprise Strategy Group (ESG), said that ESG Lab did hands-on testing to validate the performance claims. "They're really not impacting performance because they're not doing dedupe. They're doing compression," she said.
According to a December 2008 report authored by ESG's Brian Garrett with Claude Bouffard and sponsored by Storwize, ESG Lab validated that a Storwize appliance can be used to reduce primary and archive NAS capacity between 75% and 91% for real-world data sets. ESG Lab further confirmed that Storwize "not only minimizes the potential performance impact of real-time compression, it can actually be used to increase performance for real-world applications," according to the report.
Walsh said the system has no write cache, so it literally compresses the write in real-time, and puts the file with all of its attributes directly onto customers' storage systems, typically into their write caches. An I/O is confirmed back to the application before it moves onto the next write.
"We're just physically moving around and caching less physical bytes for the same data," he said.
Storwize uses standard Lempel-Ziv compression algorithms that have been around since the 1950s. "In fact, compression's really boring," Walsh said.
Storwize's technology can be used in connection with deduplication software. In that scenario, the compression is done first and the dedupe follows.
"Dedupe will pick up repetitive data. Compression will make every piece of data smaller. If you can do compression and it's performant, then you compress everything. It's lossless and deterministic," Walsh said. "But you do see applications where both make sense, and they're complementary."
Walsh said Storwize intends to keep its focus on compression. The current shipping product is NAS-only, but next year's roadmap calls for the addition of block-based storage, according to Walsh.
This was first published in December 2009