Google has posted a semi-technical explanation of the 1 September 2009 Gmail outage on the Gmail blog. To sum it up, Google took some of the Gmail servers down for maintenance. Other servers that were supposed to handle the load became overloaded and shifted the load to other servers that couldn’t handle the load and so on. The snowball grew until there were no more servers to handle the load and most people could not get to their email.
This outage illustrates a very important point about cloud computing: the cloud is not infinite. We may be under the impression as consumers of cloud computing services that the “cloud” stretches as far as the Internet goes. In reality there are a finite number of servers that with improper configuration can still all go down. This outage shows that a cloud computing service is vulnerable to the same kinds of outages that a single server or cluster of servers is vulnerable to.
Leave a Reply