Benchmarking Power-Efficient Servers

modapi writes "According to the EPA, data centers — not including Google et al. — are on track to double power consumption in the next five years, to 3% of the US energy budget. That is a lot of expensive power. Can we cut the power requirement? We could, if we had a reliable way to benchmark power consumption across architectures. Which is what JouleSort: A Balanced Energy-Efficiency Benchmark (PDF), by a team from HP and Stanford, tries to do. StorageMojo summarizes the key findings of the paper and contrasts it with the recent Google paper, Power Provisioning for a Warehouse-sized Computer (PDF). The HP/Stanford authors use the benchmark to design a power-efficient server — with a mobile processor and lots of I/O — and to consider the role of software, RAM, and power supplies in power consumption."

ummmmm

[djupedal (584558)]

When someone considers the impact, end-to-end, from carving copper oreout of the ground to throwing the out-of-date server chassis into a furnace, then I'll pay attention...maybe.

Until then, this is just marketing 101...

Not really

[WindBourne (631190)]

The lifetime costs of the chips are what you can control directly. As it is, the manufacuering of the chip (and even the systems) are going to be relatively close to each in terms of energy. The CPU/GPU is the single largest means of our being able to control energy.

Re:ummmmm

[kebes (861706)]

Your basic point, which is that we need to consider not just operating costs also manufacturing and disposal costs, is a valid one.

However the way you've worded it amounts to "since we can't account for all aspects of impact, I'm not going to worry about any aspect of impact." That's a bit extreme. Surely reducing our power consumption during the operating lifetime of our servers is a step towards greater environmental and fiscal responsibility.

Now, if you can show that the "energy saving" chips generate more pollution during production than the "normal" chips (and that this increase in pollution/energy-use/cost is greater than the savings during the lifetime operation of the chip), that's important. However I doubt that is the case. Thus, to ignore the potential advantages of power-saving measures in the data-center, simply because such measures don't address the orthogonal concerns of production impact, is silly.

Re:

[Azghoul (25786)]

... except that the Prius promises far more benefit long term than yet another diesel, by way of popularizing the idea that a car can, in fact, be driven by something other than burning dead shit.

Never misunderestimate [:)] the power of technological progress - you gotta start somewhere.

Re:

[Ryan Amos (16972)]

This is important in the IT industry for reasons other than being a tree-hugging hippie (not that there's anything wrong with that.)

More efficient, lower power servers directly relate to a cash savings on your electric bill. One server operating at 10% greater efficiency may not be a big deal, but it starts to matter when multiplied over a room of servers. Servers that use less power (generally) put off less heat, so you also save electricity because you don't have to cool them as much, and you can cram mor

Units?

[niceone (992278)]

I'm hoping the units are going to be kWh/slasdotting.

virtualization?

[stoolpigeon (454276)]

I'm not a hardware guy - but in an all hands meeting the other week, we were told that virtualization was going to save us a bunch of money on power. Our data center isn't all that big and they were talking about 2-3 thousand (US$) a month or something like that.

Re:

[eln (21727)]

Everyone is looking at virtualization for this sort of thing, and it does hold some promise. Currently, though, virtualization still comes with some very significant performance penalties. I think if virtualization can further mature, and we can get more cores and cheaper memory, we will be where we want to be.

Eventually, if RAM continues to get bigger and cheaper, more cores get packed into chips, and virtualization becomes what it is intended to be in terms of performance and stability, we will start to

Re:

[growse (928427)]

So you mean, big boxes with loads of CPUs and tonnes of memory, all connected to a huge storage system?

Sounds like IBM did a good thing keeping their mainframe business open :p

Commodity hardware was sold over big mainframes on the basis that it's much more scalable. If you want to do something else, just buy a couple of relatively cheap boxes and away you go. The thing that no-one mentioned is that it suddenly starts to cost a lot more $$$ to keep the things in power and cooled properly, so now we're se

Re:

[Znork (31774)]

"we were told that virtualization was going to save us a bunch of money on power."

Sortof. Unfortunately the ease of deployment and price reductions accomplished tend to result in a vast expansion of virtual servers instead. You're likely to end up with as much hardware except it's doing several times more than what it used to.

At least from what I've seen of virtualiztion your bill isnt going to get smaller, you're just going to get more for it. Which isnt too bad anyway.

a la simpsons

[by Anonymous Coward]

EPA Official: You've gone mad with power!
Russ Cargill: Of course I've gone mad with power! Have you ever tried going mad without power? It's boring and no one listens to you!

Efficient design

[Neil Watson (60859)]

When looking at all the servers in datacenters I often wonder how many servers are not actually needed. I do not mean redundant or due to be decommissioned servers. I mean there are to many dedicated servers. It is a common theme to have every service using a dedicated server. Is there any reason a server can not do more that just DNS? A properly built, configured and maintained server should be able to fill multiple roles thereby reducing space, power and cooling requirements.

Re:

[Calinous (985536)]

We have several web servers with IIS. Due to different software requirements (.Net 1.1 and .Net 2.0, some add-on or another), the functionality of the web servers is not guaranteed. Adding another web domain (developed with other tools) could be tricky.
      I would really like to have one server for each of the web sites running on Windows. Too bad virtualisation is out of discussion, as Windows is a big memory hog.

Re:

[samkass (174571)]

This is exactly the question answered by virtualization. Mixing lots of server processes on one OS instance makes it difficult to maintain status, monitoring, fault tolerance, etc. And before virtualization, a different OS instance required a different piece of hardware. When you hear about companies using VMware to save thousands on server power and air conditioning bills, this is where that savings is being realized.

Re:

[LWATCDR (28044)]

Well I question if the X86 is the way to go. This one size fits all mentality with CPUs needs to go. For things like web serving why not something like the SunT1 or T2. For NAS the PowerPC, ARM. and MIPS might be more power efficent.
For render farms, database servers, and HPCs the X86-64 and Power5, UltraSparc T2.

The X86-64 does a good job at about everything but it is not the best at anything. The new low power laptop cpus are not terrible but I don't think they can match the ARM, PowerPC, or Mips in the

Network Queue Systems

[Colin Smith (2679)]

Have been optimising server resource utilisation for decades.

The real problem is that most I.T. staff are either as dumb as bricks and have no idea how to make use of one or have plenty of profit to burn and just don't care.

 

Re:

[silas_moeckel (234313)]

Cramming lots of things onto one server running a single OS is generally problematic at best, it's the classic small IT shop mistake for long term reliability. Vmware can alleviate that by dedicating an OS to a function so you can do things like reboots without affecting a small pile of ancillary systems add a san and now your not adding piles of disks with each new server and can do vm level load balancing and clustering, have to replace some hardware move all the vm's off without any outage and fix the g

Don't worry

[Colin Smith (2679)]

"The Singularity" predicts that processing power will continue to increase exponentially for ever. So obviously, electricity generation will also do the same. Not a problem.

 

It's going to be interesting

[HangingChad (677530)]

If we ever run into wide-spread power availability issues. In the event of a natural or economic disaster, perhaps a series of them or we just degenerate into a civil war between political factions. No one ever imagines we could go through a near-collapse and fragmentation similar to the old Soviet Union.

We'd likely have bigger worries than whether we could keep our data centers running but it's an interesting scenario to contemplate. I honestly had no idea data centers in the US consumed that much powe

DC power

[rhaig (24891)]

everything uses DC internally. Some hardware allows for DC inputs. using DC across the board would greatly reduce cooling costs.

Re:

[Doctor Memory (6336)]

Cut and crimp positive, negative and ground, fight with heavy gauge wire, find suitable ground points, or make your own if you have to...

Nonsense. You just run the wires over to the buss bars that run between the racks, drill a new hold for each one and bolt it up. No muss, no fuss, and your wires are easy to manage. Plus, open buss bars hold serious comic potential if one of your cow-orkers likes to wear a big key ring on their belt...

The case for Smart Appliances

[maillemaker (924053)]

Years ago we heard how PCs were going to be embedded in everything from the dishwasher to the refrigerator, and I was left wondering, "why?"

Perhaps now I know.

It would be nice if I could set my house up on a "power budget", and let my appliances vie for electrical power and load-balance themselves to stay within that budget. If all appliances spoke over the in-house wiring (or perhaps wireless) and could turn themselves off or adjust their power usage that would be awesome.

You could implement something similar to this today with an X10 system or the like, but this is more of an off/on scenario, and is not based on actual power demands.

It would be great if all of my electrical things in my house could get together and say, "OK, guys, we have X amount of electricity to use today between all of us. Let's figure out, based on past usage patterns, who needs to be on and when in order to hit this budget".

Re:

[RAMMS+EIN (578166)]

``You could implement something similar to this today with an X10 system or the like...''

Dude, we've been using X11 [x.org] for some time now! X10 has been obsolete for almost exactly 20 years...

Programming languages and system architecture

[chthon (580889)]

Answer me this : how much power is lost through the use of inefficient programming languages and architectures which only emphasize processor speed, instead of balancing memory, processor and IO ?

Python, Perl and PHP all suffer from one big drawback : when you scale up you need that much extra processor power. One programming language I know (Common Lisp) offers the advantages of them, but can be compiled to near C/C++ speeds. I suppose there are others. Don't come saying that programmers are expensive. It seems that what you gain on programmers, you lose in the cost of your datacenter. I don't know how Java matches here, it probably depends upon the deployment of more recent JIT compilers.

If you see how much a process has to wait on IO, how come there are still no good solutions in providing enough IO bandwidth that the processor can use fully ? (Unless you buy a mainframe or iSeries system that is)

Just asking.

Re:

[DamonHD (794830)]

As to Java: I have just moved a rack of (Solaris) servers @670W on to a single (Linux) laptop @18W (~25W from mains, but sometimes it runs off-grid on solar PV).

http://www.earth.org.uk/low-power-laptop.html [earth.org.uk]

I actually now control the CPU-speed control with another small Java app (see update for 2007/08/20 on same page) and in particular watching it with strace() can't see the JVM doing anything that hand-crafted C wouldn't in the main loop.

In fact, the whole machine, including several Java and static Web ser