Comments on: Amazon to the rescue http://stage.vambenepe.com/archives/182 IT management in a changing IT world Fri, 29 Aug 2008 05:41:54 +0000 http://wordpress.org/?v=2.6 By: William Vambenepe http://stage.vambenepe.com/archives/182#comment-37559 William Vambenepe Thu, 27 Mar 2008 19:17:10 +0000 http://stage.vambenepe.com/archives/182#comment-37559 Thorsten: Agreed. In a way it reminds me a bit of the "~" backup file that emacs creates when you edit a file. Except now it's not a backup file with the old data, it's an entire backup machine with the old configuration. Thorsten: Agreed. In a way it reminds me a bit of the “~” backup file that emacs creates when you edit a file. Except now it’s not a backup file with the old data, it’s an entire backup machine with the old configuration.

]]> By: Thorsten http://stage.vambenepe.com/archives/182#comment-37555 Thorsten Thu, 27 Mar 2008 18:36:46 +0000 http://stage.vambenepe.com/archives/182#comment-37555 I think you're both right. If you know "it's that 2U box in rack 5" then it's not a cloud. If you know "it's in Virginia in the same (or different) location as this other box" it could still be a cloud. What's more interesting to me is the unlimited availability of boxes for short periods of time. This means that if you do a rolling upgrade of 10 app servers you don't actually upgrade, you launch 10 fresh ones and swap them in. You then retire the old ones once you're comfortable with that decision, whether 1 hour later, 1 day later, or 1 week later. If your master DB pukes, you switch to the slave and fire off a fresh slave so you're back to replicating. Then you investigate the puking at your leisure and shut the box down when done. It's this "just grab the next box" that really makes life easier, safer, and a lot more fun! I think you’re both right. If you know “it’s that 2U box in rack 5″ then it’s not a cloud. If you know “it’s in Virginia in the same (or different) location as this other box” it could still be a cloud.

What’s more interesting to me is the unlimited availability of boxes for short periods of time. This means that if you do a rolling upgrade of 10 app servers you don’t actually upgrade, you launch 10 fresh ones and swap them in. You then retire the old ones once you’re comfortable with that decision, whether 1 hour later, 1 day later, or 1 week later. If your master DB pukes, you switch to the slave and fire off a fresh slave so you’re back to replicating. Then you investigate the puking at your leisure and shut the box down when done. It’s this “just grab the next box” that really makes life easier, safer, and a lot more fun!

]]>