If there is any question whether web performance and web operations is hot, you need only look at the attendance and interest in this year’s O’Reilly Velocity conference. We’re told that the total attendance topped 1,200 — which is more than both previous years combined. In its third year, Velocity is quickly becoming a must for any web-based company. Below, I will attempt to summarize the general themes from this year’s conference, and point out the most important trends/tools/news.
This year, Velocity was divided into three tracks: “Web Performance,” “Web Operations” and “Web Operations Culture.”
No surprise, the focus on making web pages faster was the main topic of conversation. Emphasis continued to be placed on front-end performance (based on Steve Souders’ best practices), but is also shifting to the “perception” of performance. Techniques such as progressive enhancement and measuring Time-To-Interact speak to the fact that the only thing that matters is how fast the user thinks your site is, not necessarily the absolute “load time.”
To this end, the number of tools that exist to help you with improving actual and perceived performance continues to grow. We saw presentations and demos for the traditional Google Page Speed, YSlow, HTTP Watch and WebPageTest.org, plus a bevy of newer tools such as dynaTrace Ajax Edition, Google Speed Tracer, and appliances from Strangeloop that can automatically improve your performance without development effort. One unexpected trend I saw was the need for a way to measure the progress that developers and performance engineers make in improving a site’s performance (to prove to your boss and coworkers that you are having an impact) and to justify investment in performance.
We saw more correlation between performance and revenue (a major theme of Velocity 2009). There were also talks focused on the “psychology of performance” discussing the reasons behind bad performance that cause users to leave your site. Google’s Urs Holzle noted that the average page on the web loads in 4.9 seconds, but has a goal of reducing page load times across the web to 100ms (the time it takes to turn the page in a book).
Web Operations and Web Operations Culture
The news this year was that there wasn’t much news. Instead, the lessons and best practices that were presented last year are now in place, and there was a lot of evidence and examples of how effective they have been. Ideas such as devops, continuous deployment, automation and measurement are becoming commonplace in web operations departments across the industry, and there is very little reason for companies to take the “wait-and-see” approach anymore.
A lot of the web operations track was dominated by automation tools such as Chef, which allow you to manage your infrastructure (in the cloud or not) with simple command line interfaces. This included ways to track changes, handle code deployments and communicate inside your company about changes.
I presented a talk in the “Web Operations Culture” track titled “The Upside of Downtime” that you can see here. My talk focused on recommendations around preparing for eventual downtime, handling it effectively and using it as an opportunity to build stronger ties with your customers.
I’m already looking forward to next year!