Thanks for the Memory – Forgotten Secret for Data Center Productivity?

Last week while making the rounds of Oracle Open World 2011 I happened across a very interesting display at the Samsung booth. One that caught my interest almost immediately as it fit right in with the time-proven principles of Energy Logic and looked to be a play lifted right out of one of my recent blogs. Samsung was demonstrating how high-efficiency / lower voltage memory modules can significantly reduce the energy consumption of a typical server, desktop computer, or other memory intense piece of ICT kit.

Now before this starts to sound like an endorsement for Samsung’s 1.35V memory modules – a quick search on Bing, Google, and even Yahoo returned numerous other suppliers including but not limited to MicronKingston, and IDT. But I didn’t see every booth at Oracle Open World as it was a massive gathering, I got side tracked with Oracle’s America’s Cup Racer (a great showcase of what can be done with carbon fiber, though I prefer my carbon fiber on bicycles), and in full disclosure the Samsung booth was right next to Emerson’sTrellis showcase display which really made it easy and GPS free to find.

Where was I? Oh that’s right, discussing how improving your memory may be one of the easiest ICT upgrades for improving data center productivity. Meaning it provides a reduction in total energy consumption per unit of workload run in/through the data center. Now please notice that this has nothing to do with your data center’s PUE as the improvement takes place in the denominator or, as The Green Grid’s advisory council likes to point out – the focus is on the 1 (one) and not the .x (dot x) side of the PUE ratio. And frankly once your data center is in the 1.7 or below range for PUE you had better start looking at the 1 side very closely as the ROI on that side of the decimal can be significant.

From what I saw at the demo and gathered from the materials – a simple change of your servers’ memory may save upwards of 67% of the power allocated to memory or 15% of total server power. Remember the power saved from the memory modules is also saved from the DC-DC converter and power supply and is no longer rejected as heat into the data center. Total facility wide savings, depending upon your data center’s PUE and the intelligence of your cooling system, could run in the 25 to even 30% range at the building power entrance. You won’t find those kinds of savings coming so easily with anything else other than virtualization and consolidation, that is – to eliminate ICT hardware entirely.

Disclaimer – I am not your IT consultant, CIO, nor system admin. Please do check with your internal IT experts before attempting to rip-and-replace server memory! The potential savings are staggering as memory power and CPU/GPU power are at the starting point of theCascade Effect. No other actions will save you as much total energy within your data center.

For more ideas on how to save energy within your data center join The Green Grid or get involved with the Climate Savers Computing Initiative.

As always, email me with your thoughts on memory. And you can follow me on Twitter @JackPouchet


Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>


7 × two =