Google App Engine, Google's on-demand computing infrastructure service, has been growing at a rate of 25% every two months for the past six months and that has become a problem: Google has not adequately planned for App Engine's success.
Despite App Engine's increased computing footprint and its expansion across multiple Google data centers, the cloud computing service's central database, Datastore, experienced three service interruptions last month, one of which lasted 45 minutes, and is experiencing high latency again on Thursday.
The situation has become serious enough that Google has decided to stop charging for Datastore CPU usage. Google says that Datastore latency since April 1 has been about 2.5 times higher than normal.
Image Gallery: The Top 16 Google Services
|(click for larger image and for full photo gallery)|
"We want you to know we are taking the current problems with the Datastore very seriously," Google said in a blog post. "We have put other Datastore projects on hold to dedicate more people to accelerating improvements to Datastore performance, and to our datacenter configuration. We expect the Datastore may still have a few issues for the next two weeks, as we respond to the problem."
As of Wednesday, Google said it would stop charging for Datastore CPU costs until performance returned to satisfactory levels. This will be reflected starting with the May 31 bill.
The company said it will notify users via the App Engine blog seven days before it begins charging again. It also advised users not to set their Datastore CPU budgets to $0, since that would effectively prevent applications from running.
App Engine users have sought clarification from Google about its plans to address the problem.
"I would personally like more information on these Datastore growing pains," said App Engine developer Bill Edwards in a Google Groups post on Wednesday. "As a startup that has been strongly considering building solely on Google App Engine, we are very, very concerned with these instances of failure. ...If we can't depend 99.9% on GAE right now, that is fine. We will come back when you are ready. But, as a startup building a business application, we can't risk these sorts of downtimes."
Google knows it has to do better. Last month it launched a technology preview of App Engine For Business, which specifically promises 99.9% uptime though its Service Level Agreement.