AOL demos the human-free datacenter AOL has been operating a trial datacenter that runs without any on-site staff since the start of the month, and reports that the system is resilient and cuts costs. Dubbed ATC, the datacenter uses off-the-shelf, pre-racked/vendor integrated gear with open source code, is run as a 100 per cent lights out facility (no BOFH …
Simple, the vendor takes care of it. "AOL has been operating a trial datacenter that runs without any on-site staff" The vendor would not be considered on-site staff. The vendor could have someone on-site and AOL could still say that it runs without any on-site staff as the vendor is not considered staff.
I think it's more likely that the DC runs without staff say 98% of the time and then a weekly maintinace crew goes to site and swaps disk's et'al for 12 - 24 hours then theoretically moves on to the next DC in the rotation, and a serious amount of remote monitoring of the site.
With possibly a emergency crew to attend the site should the excrement hit the air con.
Either way I would be watching this with interest.
I dont want to stop progress but as an industry, IT has managed to help offshore many unrelated jobs, to the general detriment of the economy, automate more (and yes the convenience and lower unit cost is nice, but the higher national welfare budget isn't) and now we're finding means to put ourselves out of work as well, between things like this and cloud based SAAS.
All we need to do is perfect the robots to hammer the nails that can't be done over the internet and we're finally screwed, if you'll pardon the pun.
So given they have people monitoring remotely and the like then for all effect they have moved the NOC to head office. Ok so they will save on alot of costs but those costs have other area's they indirectly cover - security being one of them. Now if memory prices rise then we all know what happened during that time with RAM(sic) raid attacks. A empty no people data center would be more open to such issues than peopled ones.
A setup like this would have alot of redundancy by design, but it will only be as good as the monitoring.configuration managment/automated support scripts. Seen enough people run just at the prospect of running skulker on a nix box so nice to see things moved. But you have to cater for all exceptions otheriwse automation can compound things. Take RIM who had a core switch fail to fail over and utterly fail yet was apparently tested for such situations! Thats an exception and an example were you can't automate everything. As a rule the more people you have interacting with your system either directly or indirecly then the more possible exception your exposed too. You need people to balance out that factor. I've walked into a room and had a server smell like it was about to die (soldier burn kinda smell) - nothing on system health monitoring and was good state of the art monitoring as well and not a cheap server. Day or so later that server died, got alerted by monitoring. That example is one were human interaction was able to see a potentual problem and able to plan ahead to resolve it incase it happend and failed. But with enough spares in the racks and VM, things can be easier. if they run right but you can automate that pile of poo into a complete heap if your not careful.
But a true automated datacentre IMHO is one which nobody knows about until it retires. But anything that uses electricity or has moving parts needs love every now and then and only a BOFH can give the kind of love that is needed.
So what happened during the earthquake was that idle, unused capacity was made to do something useful, probably involving a phone call/intarweb chat to temporarily open up some latent licenses, followed by running a few scripts.
I mean, the way the spokesdroid froths you'd think it involved sparkly Star Trek replicator technology precipitating machines out of thin air to boldly supply bandwidth where no-one had done so before.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
