Go for speed

A look at a use case for Go at Iconfinder, which shaved ~40 ms off asset load times.

Transcript

1. Go for speed
2. Nick Bruun github.com/nickbruun @nickbruun bruun.co CTO @ Iconfinder Fairly carnivorous
6. Trouble in icon paradise
11. 41 ms average response time from Django application during peak hours
12. ~2,000 ms wasted per search
13. Perceivably slower
14. The internals
16. Icon ID Icon version ID Raster size https://www.iconfinder.com/icons/298255/677266/128/raster https://cdn0.iconfinder.com/data/icons/seo-smart-pack/128/ grey_new_seo2-43-128.png Public raster URL Protected raster URL
18. Constant
19. Signing requests
20. https://www.iconfinder.com/icons/298255/677266/128/raster?token=…
22. we know if the user has access right now When returning the link,
23. we know if the user has access right now When returning the link,
24. We’re okay with “right now” being an approximation
25. Time limited token
26. <token secret> = "<expiration timestamp>-<session ID>-<nonce>-<URL>"
27. <token secret> = "<expiration timestamp>-<session ID>-<nonce>-<URL>" <signature> = hmac_sha1(<token secret>, <shared secret>)
28. <token secret> = "<expiration timestamp>-<session ID>-<nonce>-<URL>" <signature> = hmac_sha1(<token secret>, <shared secret>) <token> = "<expiration timestamp>-<nonce>-base64(<signature>)"
29. But, who checks the token?
30. Enter Go The World’s least Google-able language… from Google
32. Why Go?
33. Networked service
34. Highly concurrent ~3,000 requests/second from 100 connections/core
35. Low memory footprint 14.09 MiB after 21 days
36. Low CPU usage
37. Static binary Cross-compilation = deployment heaven
38. Less decisions to make
39. Caveats
40. Structured logging
41. DIY
42. Mature runtime != Mature language
43. Conciseness 608 lines of code
44. The end result
45. 1.2 ms average response time at peak hours
46. 0.1 % CPU usage at peak hours
47. Go is awesome
48. Go is awesome for simple services
49. Go is a trade-off
50. Go is a trade-off – choose wisely
51. return