Jiffy: Open Source Performance Measurement and Instrumentation

Most tools for measuring web performance come in two flavors:

• Developer-installed tools (Firebug, Fiddler, etc.) that allow individuals to closely trace single sessions
• Third-party performance monitoring systems (Gomez, Keynote, etc.) that will hit your site occasionally and report back component-level metrics (for a fee)

Neither of these tools give you real-world information on what’s actually happening with your clients—how long are pages really taking to load, what’s the real cost of client-side execution, and what’s the impact of your loading or dependency chain. This is even more important when you don’t host all of your own assets, such as when you load ads or JavaScript from third parties, for example, and you need to monitor their performance.

Thus we built Jiffy—an end-to-end system for instrumenting your web pages, capturing client-side timings for any event that you determine, and storing and reporting on those timings. You run Jiffy yourself, so you aren’t dependent on the performance characteristics, inflexibility, or costs of third-party hosted services.

We’ve been running Jiffy at WhitePages.com since early May, with 10MM+ page views, 400MM+ requests, and 6MM+ searches/day. We’re releasing an early version of Jiffy as an open source project during Velocity, and hope others will pick it up and help us improve it.

• Provide more information on the problem and the current solutions, and why we rolled our own Provide a technical overview of Jiffy
• Show some real-world lessons we’ve had using it in our own production environment