Dell says it can save $200 per server per year on energy, but doesn't publish data to back the claim up

Dell says it can save $200 per server per year on energy, but doesn't publish data to back the claim up

Summary: Scale out. That has always been Dell's strategy when it comes adding more horsepower to a data center.

SHARE:
TOPICS: Dell
1

Scale out. That has always been Dell's strategy when it comes adding more horsepower to a data center. Just add more Intel-based boxes. Some of Dell's Intel-based boxes. Over the years, that has been the message coming from Austin, TX where Dell is based. Not surprisingly, with all of its server eggs in the Intel camp until very recently, that's all you could do if you needed to scale with Dell boxes: scale-out.  That's because Intel didn't offer much in the way of anything else that Dell could wrap some sheet metal and a hard drive around. However much Intel could scale, well, that was pretty much the limit of how much Dell could scale. Sure, Dell may have added AMD's chips to the mix, but that hasn't done much to change the company's scale out message.

But with energy costs where they are, is scale-out really the right way to go? We hear about the sorts of server farms that companies like Google are building in the Pacific Northwest where there seems to be no limit in sight to the amount an architecture like Google's can be scaled out. But sooner or later, if someone comes along with a what-if spreadsheet that proves that, between today's rising energy costs and scale-up servers that don't cost an arm and a leg to acquire, that thousands if not hundreds of thousands or millions of dollars can be spared in power-up and cooling costs (not to mention the earth-friendliness), should CFOs and CIOs be listening? In this day and age of Toyota Priuses flying out of showrooms, couldn't scaling up be the politically correct thing to do?

Well, for the time being, Dell is staying the course on scale-out but it's also looking for ways to help its customers save energy. Today, the company announced two new energy conscious servers: The PowerEdge 2950 and the PowerEdge 1950. Both are rack-mounted systems (one is a 2U, the other a 1U -- see what a U is all about) so these power saving pups are not for people buying towers or blades just yet.  

I listened into Dell's press conference today when the company's director of PowerEdge servers Jay Parker provided all the gory details for the two new servers.  But I was a bit skeptical of the company's claim to save $200 per server per year in energy costs based on some performance per watt testing that Parker said the company had done. First of all, there is no standard for doing performance per watt benchmarking and second of all, for those sorts of claims to hold up, I think a vendor really needs to transparently publish all of the benchmarking data and methodologies that led to the claim.

For example, after the press conference, I asked Parker some questions about the testing methodology and whether the tests were run when the servers were at full utilization and he said "average utilization."  For customers to better understand if they can realize the same claims that Dell cited today, they need to be able to see Dell's testing methodology and decide for themselves whether the usage scenario used in that testing maps closely to their own usage scenarios. Parker said that while the benchmark data is not public information, they will make it available to buyers upon request (so, you should request the data before buying into either of these servers). Even so, only making benchmark data available on request means that you, me, and others have to tell Dell how to find us if we want to see that data. My feeling is that if Dell can broadcast its claim to everybody, then everybody should have the right to independently verify it without having to seek Dell's permission.

In addition to hearing my line of questioning about benchmarking systems for their performance per watt and my asking Parker about scaling up versus scaling out in the context of offsetting rising energy costs, you can hear the entire press conference in today's podcast.

Using the Flash-based player above, you can stream the podcast directly to your desktop, download it, or if you're already subscribed to ZDNet's IT Matters series of podcasts, it should appear automatically on your desktop/notebook or portable audio playback device (depending on how you have your podcatcher configured).  For more information on how to tune into ZDNet's podcasts, check our How-To.

Topic: Dell

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

1 comment
Log in or register to join the discussion
  • A Slippery Slope

    Making ANY generic claims about power and cooling in relations to servers is a fool's errand. The calculated values will vary WIDELY depending not only on processor model, but front side bus speed, RAM speed and amount, as well as efficiency of power supplies and fans.

    Beware ANY "rule-of-thumb" claims made by vendors unless you know for a fact they are "worst-case-scenario" calculations. Even then, the capacity of your data center to COOL these servers will likely vary from aisle to aisle.

    No doubt, power and cooling are probably #1 in keeping IT folks up at night these days, but without critical analysis of your individual situation, you can make some very WRONG assumptions which may affect your environment for years down the road.
    JackPastor