LDNWebPerf March 2016 - Tyler McMullen

What we can learn from CDNs about Web Development, Deployment, and Performance

Transcript

1. What we can learn from CDNs about Web Development, Deployment, and Performance Tyler McMullen, CTO & Mad Scientist WebPerf Meetup, London March 1st
2. Lesson: People Use Their CDNs Wrong
3. CDNs offer a toolset • The black box approach isn’t always good • Configuration isn’t trivial – And a lot still depends on configuration • Can’t depend on the CDN to solve all your problems • Don’t exacerbate your problems!
4. http://bigqueri.es/t/sites-that-deliver-images-using-gzip-deflate-encoding/220
5. Gzipping Images • Not a very good thing for performance – Extra bytes – Extra work for the browser • But was this the Surrogate’s fault?
6. More Examples • Bad caching headers – max-age, s-maxage have a lot of power! – stale-if-error and stale-while-revalidate are rad! • Bad TCP connection management at origin • Not Gzipping (actual, compressible content) for origin fetches
7. With Great Power…
8. Lesson: Dynamic Content Is Really Interesting!
9. What Is Dynamic Content? • Stuff that’s not static! • With web traffic, generally the base HTML – Big deal because it’s blocking – And sometimes the largest object; longer download • Some AJAX • More…
11. Blocking
12. Classically, with dynamic content… Caching
13. Caching vs. Invalidation
17. We tried…
18. Dynamic Content Caching Problems • Serving stale pages – Lack of good invalidation framework • Visibility • Logging
19. CDNs and Dynamic Content • Generally, handling dynamic content has been a matter of transport – Middle mile optimizations – TCP tweaks • Some edge micro caching, but not easy • ESI
20. Actually… • Dynamic content is more cacheable than we think • Static for short periods of time • Unpredictable invalidation – Standard HTTP caching rules aren’t good enough • Event-driven Content
21. So Many Benefits! • Performance – Faster time to first byte – Faster start render – Happy users! • Offload – Less work for our servers – Less bandwidth at origin
22. What Would Make It Better? • Programmatic Invalidation – Granular – Instantaneous • Control at the edge, and not just for web pages – Real-time log files – Imagine terminating beacons at the edge!
23. Lesson: Measurement is Still Hard
24. Client-side Measurement in CDNs • Cache hit ratio – How do you test and measure? • Long tail content? • DNS and edge node selection • TTFB out of datacenter – Memory hit vs disk hit vs mid-tier hit vs miss • RUM and synthetic (Cedexis, Catchpoint, etc)
25. Let’s Test It! • 3 Objects on the same CDN (anonymous) – Cedexis object – Small image from Alexa 5000 site – Long tail object: ~40 times every 3-4 hours • Use Catchpoint last mile clients in US – Test every 15 minutes – ~11,500 total tests across all test nodes • Focus measurement on: – Connect time (TCP) – Wait time (TTFB)
26. Cedexis Object Connect (median) Wait (median) Cedexis 14ms 19ms
27. Cedexis Object Alexa 5000 Connect (median) Wait (median) Cedexis 14ms 19ms Alexa 5000 14ms 24ms
28. Cedexis Object Alexa 5000 Connect (median) Wait (median) Cedexis 14ms 19ms Alexa 5000 14ms 24ms 26%
29. Cedexis Object Long Tail Alexa 5000 Connect (median) Wait (median) Cedexis 14ms 19ms Alexa 5000 14ms 24ms Long Tail 15ms 29ms
30. Cedexis Object Long Tail Alexa 5000 Connect (median) Wait (median) Cedexis 14ms 19ms Alexa 5000 14ms 24ms Long Tail 15ms 29ms 20%
31. Cedexis Object Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 481 14ms 19ms 6741 14ms 20ms Disk 428 12ms 24ms 9626 15ms 28ms 4692 14ms 31ms Miss 1 6ms 38ms 1355 16ms 51ms 28 13ms 45ms
32. Cedexis Object Alexa 5000 Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 6741 14ms 20ms 481 14ms 19ms Disk 428 12ms 24ms 4692 14ms 31ms 9626 15ms 28ms Miss 1 6ms 38ms 28 13ms 45ms 1355 16ms 51ms
33. Cedexis Object Long Tail Alexa 5000 Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 6741 14ms 20ms 481 14ms 19ms Disk 428 12ms 24ms 4692 14ms 31ms 9626 15ms 28ms Miss 1 6ms 38ms 28 13ms 45ms 1355 16ms 51ms
34. Cedexis Object Long Tail Alexa 5000 Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 6741 14ms 20ms 481 14ms 19ms Disk 428 12ms 24ms 4692 14ms 31ms 9626 15ms 28ms Miss 1 6ms 38ms 28 13ms 45ms 1355 16ms 51ms 99.99% Mem: 96.27% Disk: 3.72%
35. Cedexis Object Long Tail Alexa 5000 Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 6741 14ms 20ms 481 14ms 19ms Disk 428 12ms 24ms 4692 14ms 31ms 9626 15ms 28ms Miss 1 6ms 38ms 28 13ms 45ms 1355 16ms 51ms 99.99% Mem: 96.27% Disk: 3.72% 99.76% Mem: 58.82% Disk: 40.94%
36. Cedexis Object Long Tail Alexa 5000 Count TCP TTFB Count TCP TTFB Count TCP TTFB Mem 11,074 14ms 19ms 6741 14ms 20ms 481 14ms 19ms Disk 428 12ms 24ms 4692 14ms 31ms 9626 15ms 28ms Miss 1 6ms 38ms 28 13ms 45ms 1355 16ms 51ms 99.99% Mem: 96.27% Disk: 3.72% 99.76% 88.17% Mem: 58.82% Disk: 40.94% Mem: 4.19% Disk: 83.98%
37. Measurement! • Not only do I care about: – Cache hit rate – Long tail – Measuring the right thing • Fetching from disk could suck! – SSDs! • Caching ≠ Caching
38. Lesson: CDNs Are Not Solved!
39. We Don’t Cache As Much As We Should! • HTML and other dynamic content • Worse cache hit rate than we think – Especially for long tail content • Mobile Apps, APIs, etc
40. Lots of Room • Making changes still sucks • Can’t take some things for granted: – DNS – Routing – TCP – SCALE! • Plus: lots of room to be creative at the edge!
41. Thank you!