This fascinating talk by Raffi Krikorian, Director of the Applications Services group at Twitter, delivers massively on technical details behind the Twitter architecture. If you’re into that kind of thing, do watch it – seriously. It’s compelling beyond belief.
We have, however, compiled a list of fun facts, to whet your appetite:
- The Twitter Firehose (containing all Tweets and events happening on the Twitter platform), which is consumed by a select few high-paying clients of Twitter, contains between 20 and 25 megabytes of data. At any given moment, on average, Twitter maintains a million such connections. That is between 20 and 25 terabytes of data they’re sending out per second!
- Each event on The Twitter Firehose is delivered to these clients within 100 – 150 milliseconds of it being registered
- In this pool of all things happening on the Twitter Platform, there are ten times more follows and unfollows between users, than there are tweets. This means users in the Twittersphere are socially active, but there are way more consumers than producers.
- 300000 distinct tweets are being delivered to distinct endpoints per second. These include SMS’s to mobile handsets, various different Twitter clients, servers receiving Twitter data.
- Twitter runs a cloud of Amazon EC2 server instances which simulate Twitter Clients, constantly tweeting, retweeting, following, unfollowing, and reading timelines, to continuously test Twitter at the highest level.
Now, go watch that talk – it’s worth it!
Posted by Adriaan Pelzer