Rearranging Deck Chairs on the Titanic

Way back when (last year), this article appeared in conjunction with discussions of soaring IT power costs.  The writer quoted EMC on the topic of power consumption and documented this curious comment:

EMC Corp., for example, has seen a surge in interest for 500 GB disk drives in its Symmetrix DMX platforms, according to Bob Wambach, director of EMC Symmetrix product marketing. The reason? A 500 GB drive offers 66% more capacity than a 300 GB version but consumes the same power in the same space footprint, he said.

Last week, EMC delivered SATA support and higher capacity drives, touting more space with the same electrical consumption footprint.  They were actually lagging behind NetApp, which had been crowing the same tune and released a paper on the subject in February 2007.

Other articles, like Chris Mellor’s in Computerworld UK, have covered additional options for reducing power consumption, like Massive Arrays of Independent Disk (MAID) and De-Duplication.  The former, MAID, is what is inside Copan Systems’ boxes (though you could do it less expensively on Zetera gear using selective SCSI “sleep” commands).  Basically, you spin disks down when you aren’t using them to save power and generate less heat.  Here is Copan’s nifty power savings calculator. 

De-Dupe is what a bunch of vendors are offering to skinny down the volume of data bits that are being stored.  You can de-dupe at the file level or at the block level, depending on the product you buy.  Either way, you compress (they hate it when you use that term) data, thereby saving space and theoretically reducing the amount of storage you need to field.  While they have jawed about this “green” impact in interviews, Data Domain seems to have spared us the inevitable white paper on the correlation between de-dupe and energy efficiency.  (NetApp did cover this in the previously cited white paper.)

As for the benefits of “thin provisioning” as a power savings technique, 3PAR has been all over this one since as early as 2006.  It is mentioned in this article, and I am sure that if I did enough digging, I would find a paper from the company some where on this topic.  Basically, their case is that we aren’t using the capacity we have efficiently.  A lot of it is reserved but not allocated — hence, blank.  By playing a shell game with this space, you can put more data onto physical spindles (until you run out of real physical capacity, which of course they assure you will not happen without plenty of advanced notice).  Reduces the need to field additional disk, thereby saving Kilowatts.

A few of the above mentioned links offer that mixing and matching these cool technologies will dramatically reduce power consumption in what is rapidly becoming the biggest power piggy in the data center:  storage. 

To his credit, Chris Mellor sneaks in a last-minute tip of the hat to tape and optical as the be-all-and-end-all of low power data consumption.  Even with all of the other technologies cited above put into some sort of rational play (assuming that vendors will work well enough together to let that happen), all of the power savings combined wouldn’t equal tape or optical, which is where the 80 percent of data that is no longer being rereferenced on disk should be parked. 

There, or in the trash can.

My question to the industry:  do we look new to you?  Do you really think throwing “value add” bling storage at a power problem is doing anything more than re-arranging deck chairs on the Titanic? 

Come on, why is PG&E giving an energy credit to companies that deploy COPAN, which promises (according to the press release) to help customers realize a 75% reduction in power consumption in their storage farm?

Release Date: April 13, 2007
Contact: PG&E News Department

PG&E Announces First of its Kind Incentive for New Energy Efficient Data Storage Technology

Program Expands PG&E Financial Incentives for Efficiency Upgrades in Data Centers

SAN FRANCISCO – Pacific Gas and Electric Company today announced the first ever utility financial incentive program to support a new high-efficiency data storage system that addresses the problem of growing energy use in data centers. The technology is called Massive Array of Idle Disks, or MAID, and it stores rarely-used data to hard disks that are normally turned off, helping customers realize 75% or more in energy savings compared to typical systems.

“Data storage growth rates for many businesses exceed 100%, so a technology that supports that rate in a more energy efficient manner will help our customers manage their costs,” explained Brad Whitcomb, vice president of customer products and services for PG&E. “By providing financial support, we hope to ramp up industry adoption of this technology.”

MAID systems manufactured by COPAN Systems of Longmont, Colorado qualify for the PG&E incentive and save PG&E customers money in direct energy costs and lower cooling system usage to help make these projects more financially feasible.

“COPAN is thrilled to be a part of the energy solution set for data center operators, and to be included in PG&E’s program portfolio,” said Roger Archibald, senior vice president of marketing and business development for COPAN. “We look forward to helping customers get a handle on data center growth, especially in terms of energy use.”

The COPAN MAID unit accomplishes energy savings in a straightforward way, but one that has not been widely adopted in the industry. Data storage needs can be characterized into different types, depending on what it is used for and how often it is accessed or changed. Data that is “persistent” – rarely changed and infrequently accessed – is saved to disks in a MAID system that is normally powered down. The information can be accessed again, but the system limits the maximum number of disks that are on at any given time.

“The energy savings, both for powering disks and cooling the unit, are certainly attractive, and from a utility standpoint, we are assured that the unit delivers energy demand reduction,” according to Mark Bramfitt, principal program manager for PG&E. “In combination, this is an exceptional set of environmental, capacity, and cost saving benefits for our customers.”

PG&E offers a comprehensive portfolio of programs and service offerings for the high tech sector, including financial incentives for customers who pursue energy efficiency projects in their data centers. Data centers can use up to one hundred times the energy per square foot of typical office space, so the energy efficiency opportunities are significant. PG&E customers in northern and central California who are interested in earning financial incentives for MAID and other data center efficiency measures must apply prior to implementing projects and can find more information about the company’s programs and service offerings at www.pge.com/hightech.

Where are the PG&E credits for all the other stuff the other vendors are shilling?  And while we are at it, where is the bonus for deploying tape or optical technology?

Guys, this all comes down to managing data better. 

Hardware is secondary. 

While I am sure that you can stall the growth of storage capacity using techniques like de-dupe (at the file level and at the block level), the real solution is to carve out the 14 TB of dirty pictures and funny videos and bittorrented music files that everyone has in their infrastructure today, right next to their undifferentiated business data.  Then focus on sorting out the stuff in DBs, file systems, ECM folders, etc. and getting into an organized archive strategy. 

This not only saves energy, but also puts your data right with the regulators.

There, I said it.  We are all dressed in our best, and ready to go down with the ship.