Skip to content

Latest commit

 

History

History
94 lines (68 loc) · 3.72 KB

trial-by-fire.md

File metadata and controls

94 lines (68 loc) · 3.72 KB
Raw

Trial by Fire: Real World Performance Optimization

Speaker: Joseph Jasinski

more than 100ms bothers users

My site is slow... where do I start?

It's not usually a quick solution.

  1. Find problems in the front-end. Possibilities: large and unoptimized payloads. Large quantity of blocking requests. Slow responses from 3rd party resources. Use chrome developer tools to analyze requests. Or you can use an external application, such as Google PageSpeed Insights, Pingdom's Speed Tools, Yslow Browser Plugin.
  2. Finding problems on the back-end. Large quantity of SQL queries, or complex queries, performing unnecessary actions in the request/response cycle. Use django-debug-toolbar to inspect queries and view other metrics. Also you can extend it with additional plugins, look for Cache Panel and Template Timings. There is also profiling middleware you can install.

Improving Performance

Front-end

  • HTML Minification, remove whitespace from your HTML. Always test your pages after installing to make sure the minification doesn't break anything. It will increase CPU time, but decrease network load.
  • Combine and compress javascript and CSS. Use django-compressor with stuff like CSSTidy, CSSmin, JSMin, Slimit, Closure Compiler, YUI. CSSmin is easy to install.
  • Optimize images -- remove metadata. Keep them below 70kb. Offer WebP to supported browsers. Use thumbnails instead of full images with something like easy-thumbnails. Use sprites when possible, especially images used frequently throughut site.
  • Resource order - put critical stuff first.
  • Lazy Loading - don't download image for the bottom of the page until the user scrolls down there. Especially useful for long web pages.
  • Use an asset CDN - serves assets from closest server, reduce load from server, but note that each alternate domain used requires a DNS lookup and a round-trip.

Back-end

  • Use "values_list" when possible so Django doesn't query all model props.
  • Don't be afraid to write raw SQL.
  • Verify you aren't running the same query twice
  • Use select_related when you know you will need related data and Django will make fewer queries. But don't use it willy-nilly.
  • prefetch_related works similarly
  • cache... how? which backend? generally use memcached, but redis works well too. low-level caching gives you the most control. prime your cache. template fragment caching is nice, and accepts context variables.

CDNs

Browsers support ~6 connectionsper hostname. You could host on multiple CDNs to maximize the number of requests the browser can do. But each host is an additional DNS lookup, so measure it.

Limiting what you do inside of a request

Always ask yourself "do I really need to do this in this moment?" and if the answer is "no", defer it (with e.g. celery or redis queue).

Headers

Set the expires and cache-control headers. You can set a far-future expires on files served by django-compressor because the url changes each time you change the static files.

Gzip

Gzip your content with the gzip middleware in Django or by setting it in your apache or nginx config file. But for now don't gzip https responses because there is a security leak.

Other

Look into Google ModPage speed (?). Database connection poolers reuse db connections. Don't use db for sessions... move to redis or memcached.

Questions

Cache invalidation for template fragment caching -- there is a package you can install to get cache keys for these so you can invalidate them. (Google for it)

You can use PIL to create sprites from user-generated image content, and create a long or wide image where all the pieces are the same height or width and you can programmatically determine the position of the correct slice.