On a official blog post Ben Treynor, vice president of engineering at Google, explained that the cause occurred after a chain overload during routine maintenance. “We underestimated the burden that some recent changes (some ironically intended to improve service availability) would promote on request routers – servers that direct queries to the appropriate Gmail server for response.”
“Around 12:30 pm Pacific time, some routers were heavily overloaded and in effect applied a message to the rest of the system: 'stop sending traffic, we're too slow!'. This load shifting to the other routers made some others too overloaded, and within minutes almost all existing routers were at their limits. As a result, people could not access Gmail via the web interface because their requests could not be routed to a Gmail server, ”he explained.
The company also reports that it will take further steps to ensure that such failures do not occur again. “We will work hard over the next few weeks to implement reliability improvements in Gmail – Gmail remains more than 99.9% available to all users and we are committed to maintaining events such as Tuesday, notable for their rarity. "