The datacenter is the new mainframe

From Greg Linden comes a link and a reference to a very interesting book by two Google engineers: The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines (PDF, 2.8Mb) by Luiz André Barroso and Urs Hölzle. This is a fascinating introduction to data center design, with useful discussions of architecture, how to do cooling and reduce power use (it turns out, for instance, that getting computers that use power proportionally to their level of use is extremely important).

I suspect that even highly experienced data center designers will find something useful here. The book is written for someone with some degree of technical expertise, but you do not need a deep background in computer science to find much here that is interesting and useful.

One of my recurring ideas (and I am by no means alone in thinking this) is that the Norwegian west coast, with its cool climate, relatively abundant hydroelectric energy and underused industrial infrastructure (we used to have lots of electrochemical and electrometallurgical plants) could be a great place to do most of Europe’s computing. Currently we sell our electric energy to Europe through power lines, which incurs a large energy loss. Moving data centers to Norway and distributing their functionality through fiberoptic cables seems a much more effective way of doing things to me, especially since that region of the country has a reasonable supply both of energy engineers and industrial workers with the skill set and discipline to run that kind of operation.

Now, if I could only find some investors…

4 thoughts on “The datacenter is the new mainframe”

  1. Torgeir Waterhouse

    This is a topic that is high on the agenda in our (IKT-Norge) agenda in our Green IT (GrønnIT) project.
    It really makes good sense for Norway to look into this…!

  2. Svenn Richard Mathisen

    Jeg har ikke sjekket dette ordentlig ut, så hvem som helst innen faget har nok bedre innsikt, men er ikke i så fall Island enda bedre skikket?
    De klarer ikke engang frakte energien sin utenlands for salg.
    “… only 20% of the conventional geothermal potential available for electricity production in Iceland, have been harnessed”
    http://www.guardian.co.uk/environment/2008/apr/22/renewableenergy.alternativeenergy
    “Renewable energy potentials in 2050 shows the following results:
    Total Demand = 6.6 TWh/y
    Total Economic Renewable Potential = 233.8 TWh/y”
    http://renewenergy.wordpress.com/transition/transition-europe/transition-countries/iceland/
    Husker ikke kilde, men har tidligere lest 65 TWh/y tidligere som potensial for Island sin del. Så tallene svinger naturlig nok mye. Men det er ingen tvil om at det er der vi finner billig overskuddsenergi.
    Til sammenligning er produksjonen i Norge oppunder 130 TWh/y. Og vi bruker opp nesten alt.

  3. Espen

    Det stemmer at vi bruker nesten alt, men en hel del går til eksport – og den kunne i stedet brukes innenlands, konverteres til bits, som så kan selges overalt gjennom fiberoptiske nettverk.
    Jeg vet det er bedrifter som gjør dette allerede, men siden dette høres så opplagt ut på overflaten, tror jeg det er verdt en studie for å finne ut hvordan det egentlig ligger an. Norge har tradisjonelt hatt konkurransefortrinn pga. vår billige fornybare energi, det bør vi ha i fremtiden også.
    Forøvrig bygger såvidt jeg vet Google et datasenter på Island. En bekymring der er muligens vulkansk aktivitet, men på den annen side ser det ikke ut til at Californiske datasentre er svært bekymret for jordskjelv…

Comments are closed.