Varnishcon 2016

Nico Stapelbroek

To celebrate the 10th anniversary of its HTTP accelerator, Varnish Software launched their yearly European conference: Varnishcon. During these two days in June the Varnish Core Developer team and user community gather together to shape up the future of Varnish Cache. As Enrise we love having available applications. Tim and Jeroen even wrote about Varnish a few years back. So now its time for me as a novice user to report on what I’ve experienced at the conference.

Although the conference was a 2-day event, I was only present on the second day of the event. The first day was a workshop day where you would take part in all types of tutorials and hands-down approaches on how to use Varnish. A training by Thijs Feryn took you through the basic concepts and use cases of Varnish. On the second day it was conference day. The day was stuffed with talks/presentations with some food and drinks during the lunch, provided by the venue at the Cristofori Concertzaal . I’ve picked my top 3 talks of that day. You can find a summary below.

Scaling Wikipedia.org with Varnish Cache

The fist talk was a pretty impressive one. Emanuele Rocca told us how Varnish cache is used at Wikipedia.org. On an average their site deals with hundred thousand requests per second. And if that’s not impressive enough: around 85% of the requests are terminated by the cache meaning it never actually hits the application. Emanuele talked about how this caching is implemented an some of the challenges they had to deal with in order to get it all working. If you’re interested you can view the talk back here.

Is the Web HTTP/2 yet?

After some talks about the current challenges and upcoming features at Varnish, Jeremy Blackburn gave an insight in the adoption rates of HTTP 2. It turns out implementing HTTP 2 is a bit more trickier than just installing and enabling it on your server. For instance: they (Jeremy Blackburn and Matteo Varvello) saw a lot of sites claiming to have support for HTTP 2 but not actually using it. If a client would send a HTTP 2 request to the server, the server would instantly downgrade the request to HTTP 1.1. This allowed Jeremy to start a discussion if it was worth implementing HTTP 2 at this time. Everyone wants backwards compatibility for older clients, but using HTTP with the tricks of HTTP 1.1. (domain sharding, image spriting, and combining resources) can negatively affect the loading speeds of your site. Resulting in a “counter productive” effect if you profile the load time of the website. All this gave Jeremy some more room to talk about how we humans interpret load times. I’ve found a similar, more summarized video of the talk here.

Content Delivery Network Evolution

The way Julien Coulon presented his cases and products was amazing. He talked about how he and his company Cedexis would solve content delivery issues by using multiple Content Delivery Networks, deploying Varnish setups in datacenters around the world and using the website visitors for reporting. The website visitors would automatically do the following tasks when visiting their site:

  • Check the health of the selected CDN’s by measuring Availability, Latency and Throughput;
  • Prioritize Content Delivery Networks on geographical location and health;
  • Report user statistics about speed back to the site

Using all this data, they created portals and dashboards for their customers to track their site performance and conversion rates. You can get an impression of how powerful this data is by looking at their live CDN and ISP outrage radar. Sadly, I could not find any sheets of info from his talk.

Conclusion

The conference was not too crowded, but certainly a good attendance for a first conference. Although some talks were too technical and in-depth for me, I’ve still learned a lot about the web, speed, varnish and how big websites deliver their content. A big thanks for Varnish Software for making this happen.

Start typing to search