Wattpad - Evolution of a Web Architecture with Amazon Web Services

34,655 views

Published on

Published in: Technology
0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
34,655
On SlideShare
0
From Embeds
0
Number of Embeds
31,074
Actions
Shares
0
Downloads
22
Comments
0
Likes
2
Embeds 0
No embeds

No notes for slide
  • \n
  • - youtube\n- upload\n- 7 million stories\n- mobile readers, website\n\n
  • - current scale\n- very social\n
  • - evolution\n- major, why\n
  • - one server\n- no caching, simple\n
  • - more complicated, lots more features\n- show you how we went from this (last slide) to this (this slide)\n
  • - mail, web ok\n- db bottleneck, own server\n
  • - web alone on ec2\n- mail\n- master for writes, slaves for reads\n- db still bottleneck, what’s hot, stories not changing\n
  • - high performance, distributed, in memory object caching system\n- in memory key-value store\n- php extension\n- cache everything\n- read-through/write-through, stampedes\n
  • - code maintainability\n- write through is data integrity\n
  • - replication LAG\n- example - delete cache\n\n
  • \n
  • - explain\n
  • - db’s humming\n- web server bottleneck, scaled vertically as much as possible\n- don’t want to split them\n
  • - caching compiled bytecode of scripts\n- apc\n- great except zero-downtime deploy problems\n- bought us a few more months\n\n
  • - one web server starting to look at a lot like this horse (donkey?)\n- need more horses\n- shared resources\n
  • - reside on web servers\n- distributed file system\n
  • - upload s3, cache hd\n- 100M req/month\n- rackspace bad\n- sessions\n- 4 web servers\n\n
  • - ELB\n- haproxy, don’t skimp on instance size\n\n
  • - horsedonkey\n- news feed\n\n
  • - facebook\n- events 150M rows\n- 500M\n- redis\n
  • - memcached but persist\n- 20k req/s\n- read counts, reading positions\n
  • - redis added\n- master doing better, slaves struggling again\n
  • - scalr, in a bad position\n- explain rds\n- concerns\n
  • - migrate\n- july 1\n- 10 slaves -> 5 slaves\n\n
  • - major things\n
  • - full text search\n- real time indexing, really fast\n- easy to scale horizontally\n
  • - row level locking\n
  • - upload story\n- worker\n- also easy to scale horiz\n
  • - this is where we’re at now\n- plus data’s stuff\n
  • - still evolving\n- next year automated all scaling processess\n
  • \n
  • Wattpad - Evolution of a Web Architecture with Amazon Web Services

    1. 1. EVOLUTION OF AWEB ARCHITECTURE WITH AMAZON WEB SERVICES Richard Poirier @richpoirier
    2. 2. EVOLUTION OF AWEB ARCHITECTURE WITH AMAZON WEB SERVICES
    3. 3. 20101 million visitors / month
    4. 4. 201210 million visitors / month
    5. 5. 1&1Web / DB / Mail
    6. 6. Memcached
    7. 7. Read Through Caching
    8. 8. Write Through Caching Replication + Caching =
    9. 9. Write Through Caching
    10. 10. Cache Stampedes
    11. 11. Op Code Caching
    12. 12. Scaling Web Servers
    13. 13. Scaling Web ServersAvatars, Backgrounds, Covers, Text Files
    14. 14. Scaling Web Servers
    15. 15. Scaling the Master Database
    16. 16. Scaling the Master Database
    17. 17. Scaling the Master Database
    18. 18. Moving to RDS
    19. 19. Moving to RDS
    20. 20. Scaling Search
    21. 21. InnoDB > MyISAM
    22. 22. Background Jobs
    23. 23. Still Not Elastic
    24. 24. Thanks!http://joind.in/7423 @richpoirier

    ×