Velocity Web Operations and Performance Day 2


The main conference track for Velocity New York 2014 kicked off on Day 2. The day began with some outstanding keynote presentations. I will highlight just a few of them (as well as some talks from the ops and performance tracks) that particularly resonated with me.

We got to hear the amazing story of Mikey Dickerson, who was working for Google and was asked to help diagnose and fix the botched launch. What he found shocked him – there was no monitoring or accountability in place. Instead, they would find out from CNN to see if/when the website went down. He came for a 3 day contract and ended up working for the government full time.

Hossein Lotfi from Verizon/EdgeCast was up next to talk about the work they do around TCP optimisation. There were some great insights around how CDN performance is tied to the origin, and he uncovered some quirks around TCP behaviour across different platforms.

Camille Fournier told us how “Cloning Yourself Isn’t an Option…”. So, what can we do then?

  • Be a multiplier
  • Train your replacement
  • Cut down on non essential work
  • Remove yourself as bottleneck
  • Don’t be a jerk

After the keynotes were finished there was some fantastic sessions from the likes of New York Times, describing their journey to making a faster despite legacy software and unique user expectations of a digital edition of a traditional newspaper.

eBay brought us a deep dive into their full web performance stack and philosophy. They have a culture of real user testing (RUM) rather than synthetic tests, which is statistically insignificant compared to actual users. They have rolled their own RUM solution to get measurements from older browsers and devices that do not support modern timing APIs.

Airbnb covered their ops journey from a small website created to find cost effective accommodation for a conference, to the behemoth it is today. They have a proud history of doing things that don’t scale – in the early days they had the founders cell phone number as the support hotline! There is a “website down” office gong that is rung when there is an outage, which thankfully is now becoming a more rare event.

Share this postShare on Facebook
0Tweet about this on Twitter
Share on LinkedIn