PUBLIC MARKS with tag performance

Node.js wasn’t selected because it is easier to scale the team with a JVM base. Node.js isn’t developed enough to have standards and best practices, a large volume of well tested code. With Scala you can use all the Java code. There’s not a lot of knowledge of how to use it in a scalable way and they target 5ms response times, 4 9s HA, 40K requests per second and some at 400K requests per second. There’s a lot in the Java ecosystem they can leverage.

Fortunately, just few days before, Opera folks came out with an experimental “style profiler” (followed by WebKit’s ticket+patch shortly after). The profiler was meant to reveal the performance of CSS selector matching, document reflow, repaint, and even document and css parsing times.

A user’s cache can only hold less than a day’s worth of browsing data: An average user browses 88 pages/day, an average page weighs 930KB, and most desktop browsers cache no more than 75MB of data. For mobile, the ratio is even worse.

As a real world example of how big the impact can be on performance, I made a test with the full HTML5 specification downloaded locally and ran the profiler. I noticed few of the most expensive selectors had ":link" and ":visited". I changed these to instead be "a:link" or "a:visited" (in this case they were equivalent in terms of which elements matched). This trivial change makes the engine use a cache, and bought the selector matching time down from about 14 seconds to about 11 seconds.

They estimate Amazon would cost them 4 times much.