3

To take the concept of automation to the most time consuming part of an IT infrastructure, rack and stack of a datacenter. What if there was a fully automated system of robots in a datacenter that was able to automatically service broken hardware, and replace disk drives and the like. The most time consuming step in deploying an IT infrastructure could be optimized. What are the caveats? Why hasn't it been done yet?

msacks
  • 49
  • 2

2 Answers2

7

Such a system would be far, far more expensive than human workers. You'd have to buy the robots, redesign racks and servers to be suitable for robotic maintenance, and you'd need a staff of folks to maintain the robots on top of all that.

ceejayoz
  • 32,910
  • 7
  • 82
  • 106
  • 11
    People to maintain the robots that maintain the servers. Some day, a great business man will decide to cut out the middle man (er, middle-robot) and just hire people to maintain the servers. His idea will be considered revolutionary. – Jason Berg Jun 27 '10 at 03:07
  • Agreed. Increasing complexity of a system is rarely going to reduce faults. IMO the best solution is to build simple redundant systems and and eliminate as many single points of failure as possible. – Keiran Holloway Jun 27 '10 at 05:28
  • I'm not convinced it would be more expensive. Until an organization makes an attempt to measure the cost. Many supply chains in manufacturing have been successful by employing robotics. A datacenter could be thought of as a another supply chain, just one that moves very slowly. If they can build a car with robots I'm sure you could fix servers in a datacenter. – msacks Jun 28 '10 at 14:13
  • ceejayoz, the following statement is not directed toward you, ErikA, sysadmin1138, Oskar Duveborn, Farseeker, Zoredache: Closing this as not a real question doesn't seem fair to me. It seems to me you just didn't like the subject matter. This is a real question, contrary to your opinion. For I will have to rely on my own media outlet for uncensored innovation. – msacks Jun 28 '10 at 14:18
  • For those who are interested in resuming this discussion, I will be hosting a new discussion forum on thebitsource.com, where innovation can flow freely and uncensored for more interesting discussion such as this. You will know when you can resume this post when a forum is up in the next day or two. – msacks Jun 28 '10 at 14:22
  • If there were likely to be significant cost savings, someone like Google or Facebook or Microsoft would've done it by now. – ceejayoz Jun 28 '10 at 19:18
2

Develop a distributed computer, plan on parts failing and have the distributed computer self-heal until the humans can come in during normal hours, get a pull list, and swap machines.

http://www.youtube.com/watch?v=zRwPSFpLX8I

ttp://www.youtube.com/watch?v=xgRWURIxgbU (though, these are gen 2 servers being shown)

No one says a failed machine needs to affect operations. And there's no reason a failed machine that doesn't affect operations needs to be replaced immediately.

karmawhore
  • 3,865
  • 18
  • 9
  • +1 Built-in resilience for teh win – Oskar Duveborn Jun 27 '10 at 07:40
  • Theres are excellent videos, thanks for that. You have a good point that a failed machine should not affect operations, but I'm not sure of it's relevance to this thread. At some point the physical hardware will need to be replaced, whether it affects operations or not. Thanks for the cool videos nonetheless! – msacks Jun 28 '10 at 14:15