REST or GraphQL? A Performance Comparative Study

Transcript

1. REST or GraphQL? A Performance Comparative Study Marcos Nazário @gustavopinto Gustavo Pinto Matheus Seabra
2. ? @gustavopinto
3. HTTP @gustavopinto REST
4. @gustavopinto
5. @gustavopinto
6. @gustavopinto
7. @gustavopinto GET https://hacker-news.firebaseio.com/v0/post/8863
8. @gustavopinto GET https://hacker-news.firebaseio.com/v0/post/8863/points
9. @gustavopinto https://hacker-news.firebaseio.com/v0/post/8863/author/ GET https://hacker-news.firebaseio.com/v0/post/8863/author
10. @gustavopinto https://hacker-news.firebaseio.com/v0/post/8863/when/ GET https://hacker-news.firebaseio.com/v0/post/8863/when
11. @gustavopinto https://hacker-news.firebaseio.com/v0/post/8863/flag POST https://hacker-news.firebaseio.com/v0/post/8863/flag
12. @gustavopinto https://hacker-news.firebaseio.com/v0/post/8863/author/hide POST https://hacker-news.firebaseio.com/v0/post/8863/hide
13. @gustavopinto GET https://hacker-news.firebaseio.com/v0/post/8863/comments
14. @gustavopinto GET https://hacker-news.firebaseio.com/v0/post/8863
15. @gustavopinto { "by": "pg", "descendants": 54, "id": 126809, "kids": [ 126822, 126823, 126917, 126993, 126824, 126934, 127072, 127217, 126889, 126875, 127535 ], "score": 47, "time": 1204403652, "title": "Poll: What would happen if News.YC had explicit support for polls?", "type": "poll" "score" : 25, "text" : "<i>or</i> HN: the Next Iteration<p>I get the impression that with Arc being released a lot of people who never had time for HN before are suddenly dropping in more often. (PG: what are the numbers on this? I'm envisioning a spike.)<p>Not to say that isn't great, but I'm wary of Diggification.<p>80% of the stuff that makes the front page is still pretty awesome, but what's in place to keep the signal/noise ratio high? Does the HN model still work as the community scales? What's in store for (++ HN)?", "time" : 1203647620, "title" : "Ask HN: The Arc Effect", "type" : “story" "text" : "Yes, ban them; I'm tired of seeing Valleywag stories on News.YC.", "time" : 1207886576, "type" : "pollopt" "parts": [ 126810, 126811, 126812 ] } GET https://hacker-news.firebaseio.com/v0/post/8863
16. @gustavopinto { "by": "pg", "descendants": 54, "id": 126809, "kids": [ 126822, 126823, 126917, 126993, 126824, 126934, 127072, 127217, 126889, 126875, 127535 ], "score": 47, "time": 1204403652, "title": "Poll: What would happen if News.YC had explicit support for polls?", "type": "poll" "score" : 25, "text" : "<i>or</i> HN: the Next Iteration<p>I get the impression that with Arc being released a lot of people who never had time for HN before are suddenly dropping in more often. (PG: what are the numbers on this? I'm envisioning a spike.)<p>Not to say that isn't great, but I'm wary of Diggification.<p>80% of the stuff that makes the front page is still pretty awesome, but what's in place to keep the signal/noise ratio high? Does the HN model still work as the community scales? What's in store for (++ HN)?", "time" : 1203647620, "title" : "Ask HN: The Arc Effect", "type" : “story" "text" : "Yes, ban them; I'm tired of seeing Valleywag stories on News.YC.", "time" : 1207886576, "type" : "pollopt" "parts": [ 126810, 126811, 126812 ] } GET https://hacker-news.firebaseio.com/v0/post/8863
17. @gustavopinto GET https://hacker-news.firebaseio.com/v0/post/8863 GET https://hacker-news.firebaseio.com/v0/post/8864 GET https://hacker-news.firebaseio.com/v0/post/8865 GET https://hacker-news.firebaseio.com/v0/post/8866
18. @gustavopinto
19. @gustavopinto GraphQL Developed by Facebook Publicly released on 2015
20. @gustavopinto GraphQL Developed by Facebook Publicly released on 2015 GraphQL allows clients to define the structure of the data required, and exactly the same structure of the data is returned from the server, therefore preventing excessively large amounts of data from being returned “ ”
21. @gustavopinto type Query {
 topStories(limit: Int): [Story] }
22. { "data": { "hn": { "topStories": [ { "title": "Monte Carlo methods Why it's a bad idea to go to the casino" }, { "title": "Public Domain Movies" }, { "title": "Agent 355" } ] } @gustavopinto type Query {
 topStories(limit: Int): [Story] }
23. Research Questions RQ: Which architectural model is the most efficient in terms of performance? @gustavopinto
24. APP1 APP2 APP3 RocketBox TechTeams MyFeed @gustavopinto
25. APP1 APP2 APP3 RocketBox TechTeams MyFeed (Dropbox simplified) (Trello simplified) (Twitter simplified) @gustavopinto
26. APP1 APP2 APP3 Rest Rest Rest RocketBox TechTeams MyFeed (Dropbox simplified) (Trello simplified) (Twitter simplified) @gustavopinto
27. APP1 APP2 APP3 Rest GraphQL Rest GraphQL Rest GraphQL RocketBox TechTeams MyFeed (Dropbox simplified) (Trello simplified) (Twitter simplified) @gustavopinto
28. APP1 APP2 APP3 Rest GraphQL Rest GraphQL Rest GraphQL RocketBox TechTeams MyFeed (Dropbox simplified) (Trello simplified) (Twitter simplified) 350 LoC 270 LoC 471 LoC 388 LoC 171 LoC 265 LoC @gustavopinto
29. APP1 APP2 APP3 Rest GraphQL Rest GraphQL Rest GraphQL RocketBox TechTeams MyFeed (Dropbox simplified) (Trello simplified) (Twitter simplified) 350 LoC 270 LoC 471 LoC 388 LoC 171 LoC 265 LoC JavaScript @gustavopinto
30. network overhead Rest GraphQL Client @gustavopinto
31. GraphQL Client Rest @gustavopinto
32. GraphQL Client Rest GraphQL Client @gustavopinto
33. GraphQL Client Rest GraphQL 1 2 3 100 requests 1000 requests 3000 requests { 15x Client 1, 10, 100 users 1, 500, 1000 users 1, 1000, 3000 users Intel Core i7, 2.20GHz, 8 cores, 16 GB memory, running Ubuntu Linux @gustavopinto
34. Size of the document @gustavopinto
35. Size of the document No reduction @gustavopinto
36. Size of the document No reduction Reduction in 100% of the cases @gustavopinto
37. Size of the document No reduction Reduction in 100% of the cases Reduction in 50% of the cases @gustavopinto
38. Size of the document No reduction Reduction in 100% of the cases Reduction in 50% of the cases @gustavopinto
39. Time per request (ms) @gustavopinto
40. Time per request (ms) @gustavopinto
41. Time per request (ms) @gustavopinto
42. Time per request (ms) @gustavopinto
43. Time per request (ms) @gustavopinto