A lot has been made of data compression in recent months, both as a means to reduce storage footprints and to introduce greater efficiencies for wide-area data networking.
But while compression is a worthwhile investment, some of the claims being bandied about are not holding up under even casual scrutiny.
One of those is that compression will help reduce overall data center energy consumption. This blog on ecogeek was meant to show how compression technology from Storwize could cut energy bills up to 95 percent, simply by reducing a 100 TB database down to 10 TB. That would be true if storage was the only energy draw in the enterprise, but as the numerous comments point out, the added hardware to accomplish the compression, plus longer disk-access times and a host of other factors will most likely balance out any storage savings.
The one area where compression could have a significant environmental impact is on the WAN. Ctrip.com, an online travel service in China, uses HTTP-based data compression from Riverbed Technology to tie two branch offices to its Shanghai headquarters. The company reports a 65 percent improvement in bandwidth utilization and a 50x improvement in application performance, which results in a direct reduction in network hardware and core CPU usage.
Still, it's extremely difficult to take compression, deduplication and other storage-efficiency claims at face value because every enterprise has a unique data environment. As InformationWeek's Art Wittmann points out in this blog, EMC claims that the dedupe technology acquired from Avamar Technologies can reduce storage requirements 50-fold. While that's not beyond the realm of possibility, it would most likely happen only with extremely large data loads, such as multiple snapshots of top-tier databases. The typical enterprise won't get close to those kinds of results.
It's also true that the compression game continues to evolve. A company called Vhayu recently launched the Velocity Squeezer system as a hardware/software hybrid solution that enables 4:1 compression with no performance penalty and enterprise-level fault tolerance. The device is built on a field-programmable gate array (FPGA) and uses a standard PCIe form factor to plug into virtually any server, which the company says provides a level of performance that can't be matched by software-only solutions.
Data compression is likely to remain a crucial technology as enterprises struggle with the increasing demands. But as for saving energy, you'd be better off deploying virtualization and low-power CPUs.