Blog Post

Once Again, See How Twitter Scales

A day after Twitter experienced its “CNN moment,” John Adams, the messaging service’s operations engineer, posted a nice slide show on how the company has scaled and the tools it uses. Entitled, “Talk Cloudy to Me” the slide show has some repeats from previous presentations by Adams, such as the chestnut about measuring everything, attacking the weakest point and then measuring again, but it also goes into some depth on some new tools Twitter is using, as well as a big push for deploying the new configuration management tools early on in the creation of a startup’s infrastructure journey.

There are plenty of details on tools and software that Twitter is using, ranging from Kestrel to Murder and Loony for the hardcore operations people out there, and some insights for business people as well, such as the slide declaring “Disk is the new Tape,” offering the insight that Web 2.0 isn’t possible without RAM, thanks to the need for instant response times as perceived by the end user. But because Flash and RAM are expensive, Twitter gets around this somewhat by using caching, using its Gizzard sharding technology for scaling MySQL data stores across multiple servers and a decision to kill queries that slow down the system. For those who want to skip the highlights, they’re taken form the final slide as seen on the right.

The full presentation is below and here’s link back to our coverage of Adam’s presentation on scaling from the first Chirp conference back in 2010. Enjoy!