Twitter had aggressively deployed Ruby for their high volume web application. As their popularity exploded, performance problems threatened Twitter's growth and their reputation. The source of the sluggishness was a mystery: was it the operating system? the Ruby runtime? the application itself? Using DTrace -- and with help from the DTrace team -- Twitter was able to discover several nasty problems in the Ruby runtime that led to massive performance gains, and solved their performance crisis.

DTrace is the dynamic tracing facility for systemic analysis, and has been designed to be used on production systems as well as during development. In this talk, two members of the DTrace team will describe the steps taken at Twitter to find and solve some of their performance issues. While the focus will mostly be on Ruby, DTrace allows users to examine applications in many languages as well as every layer of the software stack down to the lowest level operations of the kernel; DTrace is an incomparable tool for developers and administrators of any kind.