Via the excellent data center knowledge website, I came to this page with some insider information about Apple’s infrastructure and their (lack of) IT plans. While it’s all information from unidentified sources, it all rings very true, especially after seeing the problems they had at the MobileMe launch. Some highlights:
- There is no-unified IT plan vis-a-vis applications; each has their own set of servers, IT practices and release scenarios.
- Developers do testing, load testing and infrastructure planning, all of which is implemented by someone else.
- They use Oracle on Sun servers for the databases and everything has its own SAN storage. They do not use active Oracle RAC; it is all single-instance, on one box, with a secondary failover.
I mean, seriously, can they run the entire MobileMe backend database on a single Oracle database? That’s a pretty heavy server they must be using. Dave Ohara has some more background, and I just love this bit:
Trying to get everything to work in a complex service is a difficult path which makes failure a high probability. Imagine what Windows Vista had been like if Microsoft released features in phases, measuring the success of features based on the downloads and activation. Base the dev teams bonus on use metrics. It’s game changing, and flies in the face of the status quo. I like it!