Thread: True Server Efficiency

It's probably like 200w in, 198w out or something with the 2w being your computing power.

What is "computing energy" in physics terms?

Originally Posted by SynergyWorks

Surely if 200W draw produces 200W of heat... the actual computing is effectivly "free" and I was always under the impression nothing was free...?

Should I put 2KW of servers in my garage as apposed to a bar heater?

I've always found that a big tower or two is a great way to heat a room and get some "free" computing whilst you're at it

Originally Posted by SynergyWorks

Surely if 200W draw produces 200W of heat... the actual computing is effectivly "free" and I was always under the impression nothing was free...?

Should I put 2KW of servers in my garage as apposed to a bar heater?

Which was exactly my point - a processor consuming 45 watts cannot physically do anything if its 100% efficient in turning that into heat, i.e. 45 watts out..

We've been performing none scientific testing of this (don't laugh too much, but it really is funny) with a huge "bag" and enclosing a rack with a portable air-con unit and a BTU meter - we have had some interest from the local UNI or more to the point prodded a mate of a mate to assist - our biggest problem has been the isolated environment, e.g. external heat leakage from the containment, but with a coating and such we are drawing conclusions. Some chap is busily working out the efficiency and wasteage of the ACU to add into the mix. Current feelings are around 75-85% of draw is converted to heat, completely contrary to Intel and Seagate, who both claim 100% conversion rate - apparently neither of their devices is less than 100% efficient at power in to heat out - which is clearly cobblers...

Originally Posted by joeyconcrete

As a side-note, what route are you taking with that? We're currently kitted out with a load of APC SX and their inRow units which seem to work well. Hot Aisle containment would be a next possible step to improve efficiency. I noticed on one of the Gyron blogs it looked like their cold aisle was contained?

Hot aisle containment - done some extensive research on this including some excellent papers from Sun who are running hot aisle at 100-104 deg F i.e. 40 deg C. Inlet temps can go up from 18 to 23 deg and we have tested with shuttering basically leaving the front edge of the rack virtually sealed to the rack and with plexiglass doors with frames to enclose the row ends. Ok I agree its not exactly comfortable in the hot aisles when you are working, but on basic energy metering we are seeing a 30 pc reduction in cooling costs - and this is not flash cooling, it is basic DX stuff at present.

Spending a lot of time scaling this to 1850 sqr foot, which is what it is for and are even further insulating the roof area as the losses due to lower external temperature are far lower than the gains from natural conducted heat at present.

The next step is to hit natural cooling, i.e. ducting the fresh air in and out of the place, along the lines of free-cool chillers but without the cost - Again here are some excellent papers and both IBM and HP have performed tests on these lines including the loss of anticipated life span on equipment due to airborne particles and continual changes in humidity (without putting duct chillers in to drop the humidity level, which is counter-productive)

We are lucky the building has "end/outside walls" at both ends of the rack runs, and we are going to be testing a pressurised air intake and a forced extract along the rows of racks later this month - however keeping the internal building pressurised is going to be a problem we suspect without some flash control gear.

Pete, I've been going through similar fag packet ideas myself over the last 6 months. My only conclusions are to eliminate the need for any significant AC within the facility itself and takle the heat at source ie take the heat straight off the sinks and deal with it or use heat exchangers built into the rear doors of the cabinets.

Heat exchangers in the cabinets is a fantastic idea if you've got the budget of crap gemenieiei or E D ssssss or similar and I believe is technically the right solution but when you look at the ROI on these things its bloody hard to stack it up against a CRAC + freecool chiller solution, or even a dry/DX and CRAC/overheads package - yes if you are pulling 7-15kw per rack ok, I agree the spot cooling is great and if it was someone elses money who didn't give a whoot (I'm thinking .gov etc bridgage) I am damn sure it would be excellent - however our combination of ducted cold air plus aisle isolation is going to be hard to beat in terms of efficiency - We are also going to be using VRF DX kit and offering free heating for the surrounding buildings subject to them paying for the pipework as it offers a heat transfer soution - we already have every room in the building, including the loos heated all FOC with heat transfer from the DC area...

We are working on it ourselves with a lot of reading and research - we have a virgin site and time to play, hence are doing.

Oh and a background in electrical engineering plus experience in building control/hvac doesn't go amiss either

Peter it's good to see someone thinking for themselves, call it 'out of the box' or 'blue sky' but looking at a problem from a fresh perspective and not accepting some $$$ solution because it's expected of you.

I've blogged about similar issues on tinfoilhats over the last few weeks, but it my best example so far is being told that £18K, 3 year contract for 500m of fibre was the best we could get, then it was impossible for anyone to compete with a DIY solution, and now I'm sat with two quotes for a complete solution at a fraction of the price.

Originally Posted by netserve

I've blogged about similar issues on tinfoilhats over the last few weeks, but it

Never heard of it I will take a look. As I said its ok to spend $$$$$$$$ if you are expecting an ROI on it, and I actually don't have a problem doing so, however when an auction turns you up a half megawatt genny with test time only on it and no one wants it so it goes for a song, would I be the stupid one to worry about transport - I think not...

Originally Posted by PeteK

Which was exactly my point - a processor consuming 45 watts cannot physically do anything if its 100% efficient in turning that into heat, i.e. 45 watts out..

Because of conservation all energy eventually ends up as low energy photons (i.e. heat).

At the end of the universe exactly the same amount mass-energy will exist as when it was created. Eventually the system (the universe) will reach a state of maximum entropy.