Serving 30,000,000 Requests an Hour in the Cloud

Terrence Ryan

Friday 27 January 2017 from 16:00 to 17:00

Talk in English - UK at PHPBenelux Conference 2017
Track Name: C
Short URL: https://joind.in/talk/74ea4 (QR-Code (opens in new window))

Avg. Rating

You’ve written an application and now the worst thing that can happen to you has come about – People are using it! You now have load spikes to deal with. You can scale up to deal with the excess traffic, but what do you do when your load is variable? How do you do that, and stay on budget?

This session is a case study on getting an App Engine app to scale to serve 30,000,000 requests an hour. It will take you through the architecture, the code changes and the technology trade offs to get to that level of traffic.

Comments

Comments are closed.

Stefan van Essen at 20:04 on 27 Jan 2017

Pretty fun how one can learn so much from a made mistake.

Rick van der Staaij at 22:00 on 27 Jan 2017

Cool to hear your story and see what's possible with the Google app engine. Not really applicable to a real website / application though.

Steve Guns at 07:20 on 28 Jan 2017

Kind of misleading in title and description as is has little value in the real world. More of a google cloud promo to me.

Łukasz at 09:36 on 30 Jan 2017

You cannot make any comparision showing that go is faster than php 5.5 (which is not maintained anymore). It's shame that goolge isn't treat PHP seriusly. For me it's too mutch of promo of unusable product.

Terrence Ryan (Speaker) at 22:50 on 4 Feb 2017

Łukasz - terribly sorry that I was not clear enough in what I was saying. I wasn't saying that go was faster performing in general than PHP 5.5 or 7. That's not fair in the case of 5.5 (which is clearly outdated as you say) and not provable in the case of 7 (which doesn't run yet on App Engine and therefore cannot be tested.)

What I was saying that a go binary launches faster on App Engine then PHP code. This is a pretty reasonable point, as PHP is an interpreted language with a corresponding runtime and a go binary is a complied stand alone application. With a non-exhaustive search I can't find performance metrics for PHP 7 spin up times, or anything that directly compares PHP vs golang performance.

But more importantly, I was not trying to say "use Go over PHP." I thought I was pretty clear that I was saying "if you run into a performance bottleneck in PHP you might need to try another language, like C or, like I chose, Go." You give up the productivity of PHP but you get the performance of the other language.

In any case, sorry to leave such a bad impression.