[Beowulf] A Cooler Cloud: A Clever Conduit Cuts Data Centers? Cooling Needs by 90 Percent

Prentice Bisbal pbisbal at pppl.gov
Mon Jan 28 11:11:45 PST 2019


>  I suspect though that you need servers engineered to fit onto their 
> heatsinks.

This is the main crux of my previous criticism for direct-contact 
liquid-cooling solutions, the cooling piping and heat exchanger(s) can 
be retrofit to existing server designs with minimal re-engineering. And 
even for immersive cooling, you just remove the case altogether, or make 
it completely sealed to contain the coolant, both of which require 
minimal re-engineering. This solution looks like it would require major 
rethinking of how components are organized.

Prentice

On 1/28/19 12:24 PM, John Hearns wrote:
> Prentice, the website refers to Open Compute racks. "... technology 
> has been designed to fit into standard Open Compute racks".
> So yep, 19 inch racks are not being targeted here. But OCP is pretty 
> widespread.
> I would really like to find out if they can retrofit these to existing 
> kit. I suspect though that you need servers engineered to fit onto 
> their heatsinks.
> orced Physics cooling technology has been designed to fit into 
> Standard Open Compute Racks. orced Physics cooling technology has been 
> designed to fit into Standard Open Compute Racks.
>
>
> On Mon, 28 Jan 2019 at 16:48, Prentice Bisbal via Beowulf 
> <beowulf at beowulf.org <mailto:beowulf at beowulf.org>> wrote:
>
>     So I was thinking about this over the weekend (because I
>     apparently have nothing better to do with my time), and I
>     definitely think this is a non-starter due to the massive change
>     in server hardware layout to accommodate this thing. Yes, blades,
>     and twin form factor servers already required that, and they're
>     common form factors, but those form factors were just a matter of
>     shrinking or changing the layout of the motherboard but still look
>     like "traditional" layouts to the untrained eye, and they were
>     still designed with typical front-to-back air cooling in mind. I
>     feel like re-arranging the layout of components to accomodate this
>     thing is a just more of a change than the market will accept.
>
>     Just my 2 cents.
>
>     Prentice
>
>     On 1/25/19 3:56 PM, Prentice Bisbal wrote:
>>
>>     Eric,
>>
>>     I was suspecting that might be the case, but the explanations in
>>     the other articles were way too vague to be sure of that. The
>>     NextPlatform provided much better pictures. If that's the case,
>>     this thing operates like a direct-expansion (DX) refrigeration
>>     system, where the refrigerant is air and does not change state
>>     from liquid to gas, like a typical DX refrigeration system, and
>>     the induced-draft fan provides the shaft work, and those tiny
>>     channels that allegedly line up the molecules act as many tiny
>>     offices for the throttling process. Based on the pictures in the
>>     Next Platform article, here is a crude drawing of cross-section
>>     of one of these devices that I drew in Google Draw. It should
>>     help you understand what's going inside this thing:
>>
>>     https://docs.google.com/document/d/1UK94PxVlQtVSb2ns5TbCqHjPJ1vYSOmkGSeSorvHyaM/edit?usp=sharing
>>
>>     Given this design, you can only have an induced-draft fan on the
>>     outlet. A forced-draft fan on the inlet would compress the air,
>>     heating it up and negating the throttling (or Joule-Thompson)
>>     effect on the low-pressure side.
>>
>>     At the end of the day, thermodynamics still says X amount of
>>     shaft work has to be done to provide Y amount of cooling through
>>     this process, so I'm still skeptical of it, especially at scale.
>>
>>     And for those of you looking for something really boring to read
>>     rather than work, here are the related patents. I haven't read
>>     them myself.
>>
>>     https://patents.google.com/patent/US8414847
>>
>>     https://patents.google.com/patent/US8986627B2
>>
>>     https://patents.google.com/patent/US10113774B2
>>
>>     Prentice
>>     On 1/25/19 2:26 PM, Eric Moore wrote:
>>>     Actually, it looks like Joule-Thompson cooling to me (Especially
>>>     given the "Joule Force" name). You've got the air intake
>>>     (ambient), then an expansion nozzle, into a low-pressure region,
>>>     which is created by the fan at the end. So the outlet velocity
>>>     of the air (and thus it's kinetic energy) is higher than the
>>>     inlet velocity, which would lower the internal energy, and thus
>>>     the temperature. Instead the fins/nozzle/heatsink transfer heat
>>>     to the expanding gas, which exits a little above
>>>     ambient temperature. I imagine the drawback is you really need
>>>     to get rid of that high velocity hot air, and can't recirculate
>>>     it, or the kinetic energy would be converted back to thermal
>>>     energy, and mess it all up. The descriptions do all involve the
>>>     exhaust air being ducted to the outside. This article has the
>>>     most technical detail:
>>>     https://www.nextplatform.com/2018/12/04/the-leading-edge-of-air-cooled-servers-leads-to-the-edge/
>>>
>>>     On Fri, Jan 25, 2019 at 11:33 AM Prentice Bisbal via Beowulf
>>>     <beowulf at beowulf.org <mailto:beowulf at beowulf.org>> wrote:
>>>
>>>         You all know how much I like talking about heat transfer and
>>>         server cooling, so I decided to do some research on this
>>>         product:
>>>
>>>         Here's their website:
>>>
>>>         https://forcedphysics.com
>>>
>>>         and here's their YouTube channel with 5 videos:
>>>
>>>         https://www.youtube.com/channel/UClwWeahYGuNl0THWVz1Hyow/videos
>>>
>>>         This is really nothing more than an air-cooled heatsink. I'm
>>>         afraid I'm going to have to call BS on this technology for
>>>         the following reasons:
>>>
>>>         1. It still uses air as the primary cooling medium. I just
>>>         don't think air has adequate thermal conductivity or thermal
>>>         capacity to serve modern processor, no matter what you do to
>>>         it.
>>>
>>>         2. In the videos, they present highly idealized tests with
>>>         no control to use for comparison. How do I know I wouldn't
>>>         get the same results doing the same experiment but using a
>>>         similar duct fashioned out of sheet metal.
>>>
>>>         3. Using this technology means a complete redesign of your
>>>         server hardware and possibly your racks.
>>>
>>>         4. None of the information in the videos or on their website
>>>         really explains how this technology works, and what really
>>>         differentiates it from any other air-cooled heat sink. Most
>>>         people with a good invention are usually excited to tell you
>>>         how it works. Since they brag about 30 international patents
>>>         for this, there's no need to try to protect a trade secret.
>>>
>>>         5. This statement:
>>>
>>>>         The fins work like teeth in a comb, neatly orienting air
>>>>         molecules to point in the same direction and arranging them
>>>>         into columns. 
>>>
>>>         Based on my education, this statement seems to be completely
>>>         devoid of science.
>>>
>>>         This statement seems to defy the laws of physics. Last time
>>>         I checked, unless an atom or molecule is at absolute zero,
>>>         it has movement, whether it's spinning or vibrating, or
>>>         both, so how can they get air molecules to line up all in
>>>         neat little rows, where the molecules are all pointing the
>>>         same way?
>>>
>>>         This also implies very laminar flow.  As fluid velocity
>>>         increases that the diameter of the channel decreases, the
>>>         Reynolds Number increases. As the Reynold's number goes up,
>>>         turbulence increases, so mathematically, I would expect this
>>>         flow to be tubulent, and not laminar. From my classes on
>>>         heat transfer, turbulent flow around the heat transfer
>>>         surface increases heat transfer, so laminar flow in this
>>>         case wouldn't be a good thing.
>>>
>>>         Until they can provide better comparisons with real servers
>>>         in real data center environments, I'm going to classify this
>>>         as "snake oil"
>>>
>>>         https://en.wikipedia.org/wiki/Snake_oil
>>>
>>>         Prentice
>>>
>>>         On 1/24/19 3:54 PM, Chuck_Petras at selinc.com
>>>         <mailto:Chuck_Petras at selinc.com> wrote:
>>>>         Well, this is interesting.
>>>>
>>>>         "According to Forced Physics’ <https://forcedphysics.com/
>>>>         [forcedphysics.com]
>>>>         <https://urldefense.proofpoint.com/v2/url?u=https-3A__forcedphysics.com_&d=DwMFAw&c=-_uRSsrpJskZgEkGwdW-sXvhn_FXVaEGsm0EI46qilk&r=fawF3TRTwCqlaBkoLcxYCr4F4NRwCc64hmEgi9rHPpE&m=zr6lAlVphGxOQTXSElww9hGpqb9IZPik0_MN2v8Fqjs&s=lb4Hi9X8NKIYWe_e1RU3Cw4gr9Uz_B7n5pnCNY0ss3U&e=>>
>>>>         chief technology officer, David Binger, the company’s
>>>>         conductor can help a typical data center eliminate its need
>>>>         for water or refrigerants and shrink its 22-MW load by 7.72
>>>>         MW, which translates to an annual reduction of 67.6 million
>>>>         kWh. That data center could also save a total of US $45
>>>>         million a year on infrastructure, operating, and energy
>>>>         costs with the new system, according to Binger. “We are
>>>>         solving the problem that electrons create,” he said."
>>>>
>>>>         A Cooler Cloud: A Clever Conduit Cuts Data Centers’ Cooling
>>>>         Needs by 90 Percent
>>>>         https://spectrum.ieee.org/energy/environment/a-cooler-cloud-a-clever-conduit-cuts-data-centers-cooling-needs-by-90-percent
>>>>         [spectrum.ieee.org]
>>>>         <https://urldefense.proofpoint.com/v2/url?u=https-3A__spectrum.ieee.org_energy_environment_a-2Dcooler-2Dcloud-2Da-2Dclever-2Dconduit-2Dcuts-2Ddata-2Dcenters-2Dcooling-2Dneeds-2Dby-2D90-2Dpercent&d=DwMFAw&c=-_uRSsrpJskZgEkGwdW-sXvhn_FXVaEGsm0EI46qilk&r=fawF3TRTwCqlaBkoLcxYCr4F4NRwCc64hmEgi9rHPpE&m=zr6lAlVphGxOQTXSElww9hGpqb9IZPik0_MN2v8Fqjs&s=VuDTSuinKPMpF6NCztFZkSGOVo3LD7MLjroIj_sn0ao&e=>
>>>>
>>>>
>>>>
>>>>         Chuck Petras, PE**
>>>>         Schweitzer Engineering Laboratories, Inc
>>>>         Pullman, WA  99163  USA
>>>>         http://www.selinc.com <http://www.selinc.com/>
>>>>
>>>>         SEL Synchrophasors - A New View of the Power System
>>>>         <http://synchrophasor.selinc.com
>>>>         <http://synchrophasor.selinc.com/>>
>>>>
>>>>         Making Electric Power Safer, More Reliable, and More
>>>>         Economical (R)
>>>>
>>>>         ** Registered in Oregon.
>>>>
>>>>         _______________________________________________
>>>>         Beowulf mailing list,Beowulf at beowulf.org  <mailto:Beowulf at beowulf.org>  sponsored by Penguin Computing
>>>>         To change your subscription (digest mode or unsubscribe) visithttp://www.beowulf.org/mailman/listinfo/beowulf
>>>         _______________________________________________
>>>         Beowulf mailing list, Beowulf at beowulf.org
>>>         <mailto:Beowulf at beowulf.org> sponsored by Penguin Computing
>>>         To change your subscription (digest mode or unsubscribe)
>>>         visit http://www.beowulf.org/mailman/listinfo/beowulf
>>>
>     _______________________________________________
>     Beowulf mailing list, Beowulf at beowulf.org
>     <mailto:Beowulf at beowulf.org> sponsored by Penguin Computing
>     To change your subscription (digest mode or unsubscribe) visit
>     http://www.beowulf.org/mailman/listinfo/beowulf
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.beowulf.org/pipermail/beowulf/attachments/20190128/448cd369/attachment-0001.html>


More information about the Beowulf mailing list