the Rackspace outage — could it happen to you?
The blogosphere is abuzz over Rackspace’s major outage last night. Apparently this is the first major outage in many years for Rackspace, a data center provider famous for its promise of “100% uptime” and “fanatical support”. Lots of Internet services pay the Rackspace premium (typically 2x the cost of competitors) for that assurance.
Last night, Rackspace was down for three hours. Whoops! Turns out, there’s no such thing as 100% uptime. There’s always some unforseen circumstance that will get you. In Rackspace’s case, it was a truck driver who crashed into their Dallas power transformer.
As a service provider, one of your first obligations is to identify and eliminate “single points of failure” — any one part of your operations that could knock you out. Even major service providers like Craigslist, MySpace, Yahoo, and eBay get tripped up occasionally (especially YouTube, which is noted for frequent outages). One of the more common mistakes is relying on a single data center for any non-redundant part of your service architecture. High performance requires that you factor in minor acts of God, too.
At SilverDock, we’ve built a fully redundant, fault-tolerant, widely distributed application infrastructure. If one server loses power, others will take over. If one database goes down, the others will jump in. If an entire city gets knocked off the Internet, our servers in other cities will shoulder the load. And it all happens automatically.
We can’t promise that nothing will ever bring us down. But we can promise that it won’t be single data center failure due to a power failure in San Francisco, a tornado in Iowa, or even a truck driver in Dallas.