Twitter has almost become as well-known for its outages
than for its short message service. But when was the last time you heard that Twitter was down? Not for awhile, right?
Which makes me think that the San Francisco, Calif., based micro-blogging site stability problems may now be something of the past.
“We made a lot of progress. I’d like to say that we’re out of the woods and we expect massive growth over the next year. But we were in fighter fighting-mode on a daily basis and now we’re not, we’re building for the future,” said Twitter co-founder and CEO, Evan Williams at last night’s Churchill Club event in San Francisco.
Now nothing goes out the door without proper testing
After the event, I chatted with Twitter’s operations engineer John Adams – who joined the startup last July – about the improvements he made to the new Twitter architecture.
“The issue was not about using a programming language like Ruby on Rails – that we still use along with Scala and some C language – but about the internal developing process. Since I joined last July, we found/fixed the bottlenecks of the application and improved caching. For the most part operation is very metrics driven. We don’t do anything until we have science, until we’ve got numbers. Now nothing goes out the door without proper testing and a fairly large amount of real thinking behind it”, explained Adams.
And Twitter’s new stable architecture proved itself during elections’ night when it handled 10 times the traffic than normal!
Jean-Baptiste Su
Mark Boslet
