Green Storage Index Heads for the Green (24-Aug-07)

From Lauraibm

Revision as of 11:30, 7 September 2007 by Laura (Talk | contribs)
(diff) ←Older revision | view current revision (diff) | Newer revision→ (diff)

Contents

MI Summary

Full article: Green Storage Index Heads for the Green (24-Aug-07)

The Storage Performance Council (SPC) will produce a standardised metric for the power efficiency of storage gear. It has been proposed that the metric will involve a financial element, because customers need to see how much they are spending in relation to power consumption and performance.

The requirement for the metric becomes increasingly urgent as data centres are reaching the limits of their cooling and power supply, and customers are being faced with the problem of sticking within their kW budgets.

Text of Article

By Tim Stammers

Whatever the Storage Performance Council's proposed green storage index eventually looks like, it is likely to include cost or purchase price as a parameter.

The SPC has already declared its intention to produce some standardized metric for the power efficiency of storage gear, but has not yet said what that might look like.

Now the multi-vendor body says that it will not simply be adding an extra column of power consumption figures to its existing test, but could well create some type of formula that would mix dollars in with kilowatts.

SPC administrator Walter Baker said yesterday:"I'm not sure what the metric will look like, but price may well appear in there."

Sun's principal benchmarking manager Leah Schoeb is also an SPC council and steering committee member, and was adamant that money needs to be in the mix.

"Customers want to see how much they're spending in relation to power consumption and performance," she said.

Schoeb said that this is why in 2005 Sun began selling the concept of a server performance and efficiency measure it calls SWaP, calculated as performance divided by the multiple of size and power.

That SWaP measure is very similar to disk array maker Pillar Data Systems' "storage efficiency quotient," a measure that Pillar began promoting this year and which simply adds storage capacity to the top line -- making it capacity times performance divided by power and size.

"These independent ideas are wonderful, but we need to agree on a common industry method for measuring storage efficiency," said Parris.

Parris said that the SPC might be able to present a draft efficiency index at the CMG -- Computer Measurement Group -- annual conference this December.

Pillar CEO Mike Workman is not entirely sanguine about using the SPC tests as a basis for an efficiency benchmark. That is because the disk arrays built for SPC benchmark tests do not always represent the real world, and sometimes are racing machines fitted with very large numbers of short-stroked disk drives in order to boost performance.

"Vendors use hundreds of spindles for those tests, because the whole purpose is to win out against each other. That's got nothing to do with energy consumption or heat generation," Workman said.

SPC administrator Walter Baker admitted that some tests had been completed on machines with very large numbers of drives. But among other defenses, he pointed out that the practice is very transparent to customers, because of the way that SPC results are published. Using a large number of drives bumps up one of the major SPC indexes, which reports on purchase cost for a given performance. The practice would also have the same effect on power consumption.

But short-stroking was constrained in the second benchmark test that the SPC issued, which was called SPC-2. "The only regret I have for SPC-1 is that we didn't put that [short-stroking] limit in there, like we did for SPC-2." Baker said, before suggesting that SPC-1 might be amended in future to include such a limit.

Our View: Benchmark formulas for power efficiency might not seem hugely important, but at present storage buyers have little to go on when trying to judge the power consumption -- and hence also the heat output -- of storage devices.

Power consumption numbers given in vendors' specification sheets are usually hugely conservative overestimates of real consumption, and are really only circuit breaker guidelines. As more and more data centers and server rooms reach the limits of cooling or power supply, that is a major problem for customers faced with kW and BTU budgets.

To be meaningful, power consumption has to be measured at some standardized load or performance. The only storage performance metric in existence are those created by the SPC.

  • Source: [Computergram]

For an overview on the topic(s), see also

Personal tools