I've been playing with Amazon's S3 for a while, but tonight I had a blinding flash of the obvious. To get higher levels of bandwidth, uptime, and reliability for Feed Digest (although they're pretty good already), all I need to do is render digests every 5/10/15/whatever minutes, push them to S3, and make users grab those versions. That way Amazon deals with the heavy bandwidth traffic, and the FeedDigest servers just deal with rendering and pushing digests to S3. I've no idea why this didn't occur to me sooner, especially since I have been planning a 'push' system like this for a while now.
Request levels are currently in the 100-150 million per month range, so I'm planning to test this system with a few beta testers first, but if S3 can stand the heat, this could be a great way to push FeedDigest's capacity up into the stratosphere. I need to see capacity of 1 billion requests per month next (peanuts, really, as Google dishes up over 30 billion per month, and FeedBurner about 8 billion per month).
Hey Peter,
It would be good if you could keep us updated on how the S3 trial works out. I've got a couple of Ruby on Rails projects lined up that I'm considering using S3 for - would love to get some feedback on reliability, performance etc.
Posted by: Ben at November 30, 2006 10:42 AMWill do! I'll be posting on this blog about it as I go.
Posted by: Peter Cooper at November 30, 2006 11:51 AMReturn to the homepage.
Privacy Policy