Talks #67 - Bogdan Radulescu - Scalable and redundant infrastructures

Having your website or application go viral it's probably the thing many of us dreamed at least once. Even though it's one of the coolest thing that could happen is actually quite possible enough traffic would bring it down.
It's clear for everyone that having your website go down when you have the most people online is just not good. In order to avoid this and achieve scalability we are going to have a look at how and when we can scale different layers of technology.

Transcript

1. Scalable server setups Bogdan Rădulescu
2. What is a scalable system? Ideally: A system which can grow proportionally Real life: An inexpensive solution to grow enough
3. How do we approach this? Identify the requirements Identify the problems Design a setup Optimize
4. The requirements Handle flash traffic at a periodic interval. Have a product which can grow 10x in one year.
5. Frequent problems DISK CPU Memory
6. Layers Load Balancer Application Database
7. The database Performance tuning: innodb_buffer_pool_size innodb_flush_log_at_trx_commit
8. The database
9. The database
10. Apache httpd + mod_* Very well known and compatible MPM (prefork, worker, event) Ability to process dynamic content .htacess NginX + FastCGI Very fast and low memory usage Event based Uses external processor (FPM) The web servers
11. The web servers <IfModule mpm_prefork_module> StartServers 3 MinSpareServers 3 MaxSpareServers 5 MaxClients 30 MaxRequestsPerChild 0 </IfModule> <IfModule mpm_prefork_module> StartServers 10 MinSpareServers 10 MaxSpareServers 25 MaxClients 200 MaxRequestsPerChild 10000 </IfModule>
12. The web servers
13. The web servers
14. The overview
15. Disks
16. Horizontal vs. Vertical
17. Graphs
18. The Wikipedia Setup
19. Bogdan Radulescu <bogdan@nimble.net> Q & A