Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
IT Science

Managing the Most Remote Data Center In the World 98

Posted by kdawson
from the talk-about-latency dept.
blackbearnh writes "Imagine that your data center was in the most geographically remote location in the world. Now imagine that you can only get to it 4 months of the year. Just for fun, add in some of the most extreme weather conditions in the world. That's the challenge that faces John Jacobsen, one of the people responsible for making sure that the data from the IceCube Neutrino Observatory makes it all the way from the South Pole to researchers across the world. In an interview recorded at OSCON, Jacobsen talks about the problems that he has to face (video), which includes (surprisingly) keeping the data center cool. If you're ever griped because you had to haul yourself across town in the middle of the night to fix a server crash, this interview should put things in perspective."
This discussion has been archived. No new comments can be posted.

Managing the Most Remote Data Center In the World

Comments Filter:
  • by Drakkenmensch (1255800) on Friday July 23, 2010 @07:50AM (#33001826)
    ... to sell air conditionning in the south pole?
    • by elrous0 (869638) *
      Wouldn't it be cheaper just to put in a window?
      • Probably doesn't help having your data center full of snow..

      • by Drakkenmensch (1255800) on Friday July 23, 2010 @10:39AM (#33003648)
        But then a bunch of penguins could sneak in and replace all their copies of Windows 2008 server with Linux. Wait, this plan actually sounds better and better every second...
        • by kievit (303920)

          Actually practically all IceCube servers at the Pole, running the data acquisistion, processing & filtering are running linux. So those penguins would not have much to do, except join the party. :-)

          (And a nitpick: IceCube is actually at the geographic South Pole, too far away from the Antarctic coast for any penguin to reach it.)

  • by Anonymous Coward

    ... to the Wookie who was interviewing him.

  • by fuzzyfuzzyfungus (1223518) on Friday July 23, 2010 @08:01AM (#33001918) Journal
    The video makes it impossible to tell if this guy is the real thing, or if The Thing has had a change to catch him in the cold isle and duplicate him. I fear to imagine what it would be capable of once it uses the base's internet connection to discover tentacle-rape hentai...
  • by Iskender (1040286) on Friday July 23, 2010 @08:08AM (#33001966)

    The South Pole is the remotest by many standards in kilometres. However, I recall some recent research which came to the conclusion that parts of the Himalayas are the remotest on Earth. At least some parts of the year you can basically just fly to the South Pole. Not so much with the worst parts of the Himalayas - I seem to recall a minimum travel time of one or two weeks.

    There was an article on the research on the BBC site about this, but it's fiendishly hard to find. Plus points to anyone who can dig it up.

    Oh and I should avoid sounding cynical and say that the stuff in the article is certainly a cool challenge. It's still a tricky location compared to 95% of all other land, and I'd love to work on problem-solving like that myself.

    • by mc1138 (718275) on Friday July 23, 2010 @08:11AM (#33001984) Homepage
      Yes but last I checked there was no data center there... The article says the most remote data center on Earth, not the most remote spot. Plus I might put the bottom of certain parts of the ocean at even more remote than the Himalayas, as there are spots down there no one's ever reached.
    • Re: (Score:3, Interesting)

      by Anonymous Coward

      Agreed, this is not remote because you can if you need to take a vehicle all the way there. Plane/Boat/snow machine.. Top of Everest is remote as it takes a week of hiking to get there.

      Me I have a remote "datacenter" to manage.. It's atop a 200 foot tower in northern Michigan... it's at least a 4 hour drive + 1/2 hour hike and then you have to climb 200 feet vertical with all your gear.

      Luckily it has not needed to be touched for 4 years. my remote administration over ham radio frequencies has worked

      • by drsmithy (35869)

        I'll give this guy the Uber geek crown if he adds a 200 foot vertical climb to his trip to the "datacenter". Because kids, that alone is a major PITA... nothing like screwing around with a laptop 200 feet up, in wind and having the damn thing sway about 2 feet back and forth while you work. Oh... dont drop anything... during setup when the RF guys screwed up and cut my comms wire to the base of the tower I dropped a laptop from 200 feet... Even Panasonic toughbooks dont survive that fall.

        Why isn't the sys

        • Re: (Score:3, Informative)

          by Muad'Dave (255648)

          Attenuation [sss-mag.com], plain and simple. At 145MHz, the attenuation of 200' of ok coax (LMR-400 @ $0.86/ft) is almost exactly 3dB - that's 50% of your transmitted power wasted heating the coax. Additionally that makes your receiver 3dB more deaf to weak signals.

          At 450MHz, it's even worse. Attenuation is 5.4dB, which steals about 71.3% of your signal.

          Stepping up to 1/2" Andrews DF4-50A heliax, you find the price jumps to $1.69/ft but the attenuation at 450MHz drops to only 2.9dB for 200'. We're still losing 50% of our

          • by X0563511 (793323)

            So... what's wrong with fiber?

            • by Muad'Dave (255648)

              It's very hard to transmit Radio Frequency energy over fiber. :-)

              You could use fiber for the control and data signals, but then you'd still have to run power up for the transmitter, and your baseband signal would also have to be sent up that way. Audio could be modulated directly onto the fiber or digitized, but you'd still need an awful lot of (custom?) hardware up the pole to do the light->electronic->light translations. With all that up the pole, why not stick it all up there? That way you don't ne

            • by Macrat (638047)

              So... what's wrong with fiber?

              Most people just don't get enough fiber.

      • by dargaud (518470) <[ten.duagradg] [ta] [2todhsals]> on Friday July 23, 2010 @09:45AM (#33002998) Homepage

        I'll give this guy the Uber geek crown if he adds a 200 foot vertical climb to his trip to the "datacenter"

        Well, let's see if I can take that crown: I managed a bunch of experiments (and their associated computers and comm systems) at Dome C [gdargaud.net] in 2005. It's higher than South Pole. And colder to boots (we had -78C). And some instruments were atop a 100ft tower (now raised to 200ft) were it was windy as hell in addition to being as cold as stated. In winter it was physically next to impossible to climb: hard-packed snow covered the scale (you had to clean every step), your breath would freeze your clothing solid around your head, your glasses would fog in a few seconds turning you blind, and if you exposed a blip of skin it would feel like a knife went through it immediately. Gosh, I miss that place: the view was fantastic.

        • by Muad'Dave (255648)

          can you say "Winch"? I'd submit to being hauled up on a cable to the top of the tower before going through all that! I've done plenty of tower work, but always in safe, warm (sometimes hot), low-wind conditions. Rohn 25 towers are scary to climb - you look down and you can't see the tower because your feet block the view. Towers look solid from the ground, but boy do they sway when you're 125' in the air!

          • by Carnildo (712617)

            -78 C? I don't think I'd trust anything with moving parts in those conditions. Materials don't behave the way you expect in extreme cold: steel cables become brittle, lubricants freeze and shatter, thermal contraction messes with tolerances, and everything gets coated with ice.

            • by dargaud (518470)

              -78 C? I don't think I'd trust anything with moving parts in those conditions. Materials don't behave the way you expect in extreme cold: steel cables become brittle, lubricants freeze and shatter, thermal contraction messes with tolerances, and everything gets coated with ice.

              Correct. We stopped using snowmachines below -55C. And the only mechanical thing we kept using outdoors was the (Caterpilar) loader for the snow melter (to produce water). Even the soles of our special shoes would turn hard as rock below that. Sorry about the bad link in the previous post [gdargaud.net].

        • by X0563511 (793323)

          [quote]Ahem, did you get lost ?
          404 Error — Not found[/quote]

        • by StikyPad (445176)

          Fixed that for you. [gdargaud.net]

      • by tehcyder (746570)

        Top of Everest is remote as it takes a week of hiking to get there

        Couldn't you just fly up in a helicopter and jump out somewhere near the top?

        No, I have no idea what I'm talking about.

    • Having done so several times, yes, between about late-October/early-November and mid-February, you can "just fly" to the South Pole. The trip can take as little as 5 days, but 8-10 days is more ordinary. I've been around for periods in the summer where nothing came and went from McMurdo (the logistics hub at the coast, and one node on the trip) for 10 days in a row (and that's not the record).

      So for 1/3 of the year, you can get on a succession of airplanes and, weather permitting, get to the Pole in 1-2 w

    • by evilviper (135110)

      I recall some recent research which came to the conclusion that parts of the Himalayas are the remotest on Earth. At least some parts of the year you can basically just fly to the South Pole.

      Parts of the Himalayas may be difficult to physically reach, but they're not as bad off in many ways. Numerous orbiting communications satellites, for one thing, work everywhere on the planet EXCEPT for the poles.

    • by sznupi (719324)

      Since a helicopter landed on the Everest few years back, I imagine you can also basically just fly to the worst parts of the Himalayas, if you really want to.

    • Did you miss the headline? Is there a data center in the Himalayas? If not, then it's not in the running for "most remote data center".

      Besides, a location that you can't even get near for half the year seems to win.

    • http://www.visualcomplexity.com/vc/project_details.cfm?id=672&index=672&domain [visualcomplexity.com]
      Search terms: remotest travel time week bbc - you find it on the second page. How do I receive my Google-Fu-points? ;)
  • Space probes (Score:5, Insightful)

    by Honken (665599) on Friday July 23, 2010 @08:11AM (#33001990)
    What I find even more impressive is how NASA, ESA and others manages space probes I think, that's really extreme conditions in every way. Often huge communication delays and poor bandwidth, absolutely no chance of at least eventually fixing a problem on-site, hardware constantly being subjected to intense radiation and extreme temperature differences. Imagine that rather unpleasant feeling you get when you reboot a remote server and you know you won't be able to go on-site any time soon to fix it if you did something wrong, then take that feeling and add the fact that you can _never_ fix it, that it costs millions or even billions to send it there, that lots of valuable science might be lost or never take place, and that you'll be guaranteed to read about your mistake in the news the following day. I guess it calls for rather extreme levels of testing before doing any changes at all.
    • Plus, having all the usual cooling methods not work at all would be a bit of a downer(ok, yeah, normal servers do lose some tiny amount of energy by radiation, so I guess that counts).

      No conductive cooling, you are floating in the depths of space, surrounded by nothing.

      No convection. There isn't any atmosphere, nor any gravity(of use, obviously gravitation is universal)

      Even in sealed capsules with humans, forced air just moves the problem around, there isn't nearly enough air to treat it as an arbi
    • Re:Space probes (Score:5, Interesting)

      by Cornwallis (1188489) on Friday July 23, 2010 @08:28AM (#33002122)

      I remember when the LEMs took off from the moon on the later missions and some comm guy at NASA was able to track the liftoff with the video camera left on the moon. The idea that he was able to *anticipate* the liftoff and ascent and remotely track still stands out as one of the all time cool things to watch.

    • Re:Space probes (Score:5, Informative)

      by dargaud (518470) <[ten.duagradg] [ta] [2todhsals]> on Friday July 23, 2010 @09:53AM (#33003100) Homepage

      What I find even more impressive is how NASA, ESA and others manages space probes I think, that's really extreme conditions in every way.

      Antarctica can be meaner in several ways: - you don't have a direct line of communication with the rest of the world (space probes do). Hell, you can't even have a direct comm with geosync satellites. - water ! Take thin crystals of ice, add lots of wind and you end up with water deep inside even sealed boxes; hence shorts and very quick rusting of components. - temperature changes. In space you surround your satellite with some heat conductive sheets and the temp basically never changes (unless you go into the earth shadow). In antarctica you can have -80C in winter, -10 in summer. To say nothing that the cold has unforeseen effects on materials (dielectric changes, materials becoming brittle...) - unstable power: the power comes from big diesel generators and is shared between experiments, people, etc... It goes out, the temp of the room where your computer is falls to -60 in 15 minutes. Power comes back, computer tries to boot. Bye, bye hard drive. - budget: experiments for Antarctica have much less than 1/10 the budget of equivalent space experiments. And most of it is eaten by logistics. So you end up with standard computers and a few hack and a guy standing nearby (me [gdargaud.net]) to kick it if necessary.

      • Mod +1 Informative please; the link is informative as well.
      • by sznupi (719324)

        Should be really fun on a mission to Titan - in many ways, worst of both worlds (just, when firmly in the atmosphere, with 100C lower temperature and possibly greater rate of heat loss due to higher density of the atmo...OTOH, that higher density and lower forces involved in buoyancy, might slightly help against convective heat loss? At least the temperatures are quite stable...)

        But nevermind machines, this one will be relatively easy (one our probe survived already, before its batteries got exhausted) - I

        • by dargaud (518470)

          I wonder, do you have anybody at hand there who is quite into survival gear and would find it entertaining to contemplate what would be needed for a human to survive in such environment? :)

          Well, (raises hand), I have an idea... I'm into mountain climbing as well, so I can combine polar experience, mountain experience and hackery. The longest walk I took at -75C was about 4 hours long (the details are somewhere on my site) . As long as you have a warm place to go back to, you are fine, but for instance we couldn't even drink at those temps, so we came back parched. At -100C a helmet with adequate ventilation to avoid icing would be a good idea. As for clothing and footwear, I don't think heati

  • by mseeger (40923) on Friday July 23, 2010 @08:16AM (#33002034)

    I don't envy someone with a job like that. It is already very difficult to serve locations less remote. E.g. to offer a SLA for a network which spans over 30 locations in the world, we have to make sure that spare parts arrive around the world within a well defined time frame. We don't want anything fancy: a week would be finde by us. But i haven't found any transportation company that guarantees delivery on site in 3rd world countries (big cities) within that time frame. All make exceptions like "customs", which doesn't help me.

    CU, Martin

    • Re: (Score:3, Interesting)

      by i.r.id10t (595143)

      Or just blow the budget and have a spare set of equipment at each location. When something dies, take out the replacement that is already there, then worry about shipping a new replacement.

      • by mseeger (40923) on Friday July 23, 2010 @08:36AM (#33002208)

        Or just blow the budget and have a spare set of equipment at each location. When something dies, take out the replacement that is already there, then worry about shipping a new replacement.

        According to the customer, this has been tried and failed... Unused spare equipment has a tendency to vanish.

        CU, Martin

    • by pgrb (121829)

      On site spares. Make sure you have (tested and shown to be working) spares for everything, and more than one for critical components and ones that are more likely to fail while you are restocking your spares. Engineers are easier to fly around than kit. Yes, it is expensive, and yes, it's a pain to have to store and keep track of the spares inventory (which always seems to 'go walkies'), but if you have a demanding SLA, it's the only way to go.

      • by mseeger (40923) on Friday July 23, 2010 @08:41AM (#33002250)

        Yes, it is expensive, and yes, it's a pain to have to store and keep track of the spares inventory (which always seems to 'go walkies'), but if you have a demanding SLA, it's the only way to go.

        Expensive is not the problem, getting the walkies is... According to customer, they experience a strange "honor code" in third world countries... Operative systems seem to be quite safe, but any spare equipment is fair game (Africa is the biggest problem in this regard). Trying to fool the people by faking spare systems to be operative was also not successful.

        CU, Martin

        • According to customer, they experience a strange "honor code" in third world countries... Operative systems seem to be quite safe, but any spare equipment is fair game (Africa is the biggest problem in this regard). Trying to fool the people by faking spare systems to be operative was also not successful.

          I was going to suggest something like this [wikipedia.org] but of course that doesn't solve this particular challenge. How does operational redundancy work? Are live failover/load balanced systems covered by the honor code?

          • by mseeger (40923)

            Are live failover/load balanced systems covered by the honor code?

            It seems to be like this: at a power failure the systems will be turned off. Once the power is restored, they turn the systems one by one. Once everything works, all systems that are still powered off are considered "as spare".

            CU, Martin

            • Wow. I guess that rules out RAID configurations too (aside from RAID-0).

            • It seems to be like this: at a power failure the systems will be turned off. Once the power is restored, they turn the systems one by one. Once everything works, all systems that are still powered off are considered "as spare".

              What if you ensured that something live is running on each server, even if the server would otherwise do nothing more than standby in the event of a failure? Something like Nagios would tell them whether or not everything is operational and in the event the server went, these "special" services could only be setup remotely by specific highly trained Western engineers. To decoy the decoys, make sure a few cheap-ass spares get setup at the same time without any of these special services and let them go miss

        • Would it be possible to set up a seperate storage location close to where the equipment is used but with much tighter rules on who has access?

          • by mseeger (40923)
            Has been tried, only increased the amount of damage done.... The idea to ship it was not done lightly, it's last resort...
        • by LWATCDR (28044)

          I can see the logic of that honor system.
          If you are resourced staved then seeing a spare system just sitting there seems like a huge waste.
          And to them it is.
          Maybe the solution would be to pay the people there such a high wage that value their job. If something walks you fire them.

  • by slick7 (1703596)
    So, is the word "so" so, so?
  • If you're ever gripped = If you've ever griped?

    Otherwise, I think "gripped" pretty much describes the state of most of the posters here, especially if they get to write about Apple!

  • Just open a window!

  • I think just this data center and the observatory is enough to bring down the south pole ice caps... If this is taking place at the south pole, shouldn't this [nasa.gov] be concerning? Also, most of the IT staff can manage the center remotely with a minimum number of people required at base camp. With regards to power, I believe geothermal is a possibility but I don't know how that would interfere with their observations.
  • Couple that with some good planning and it shouldn't be a big deal to remotely manage it for months on end. Once a year you haul out upgrades and replacements.

    Virtualization is your friend.

It was kinda like stuffing the wrong card in a computer, when you're stickin' those artificial stimulants in your arm. -- Dion, noted computer scientist

Working...