Server modeling with mysql
This document discusses server modeling with MySQL for scalability and consistency issues. It covers: - The easy problem of a single web server and database layer bottleneck which can be solved by adding more web servers. - Harder problems of database layer bottlenecks which incorrect solutions like adding more web servers will not solve, requiring database partitioning or other techniques. - Common solutions like master-slave replication where one database is the master and updates are replicated to slave databases which can balance reads. This allows for high reads but low consistency. - More advanced solutions like two phase commit and semi-sync replication in MySQL to provide higher consistency between master and slave databases during writes. - MySQL Cluster which stores data
Server modeling with mysql
- 1. Server Modeling with Mysql charsyam@naver.com
- 2. Never Ending Problem
- 4. Scalability
- 5. Problem Client Group Web Server Web Server Data Layer
- 6. Easy Problem Client Group bottleneck Web Server Web Server Data Layer
- 7. Solution Client Group Web Server Web Server Web Server Web Server Scale up Data Layer Scale out Just Add Server
- 8. Hard Problem Client Group Web Server Web Server bottleneck Data Layer
- 9. Incorrect Solution Client Group Web Server Web Server Web Server Web Server Bottleneck Harder! Data Layer
- 10. Scale up - Solution Client Group Web Server Web Server Need Money Data Layer
- 11. Scale Out - Solution Client Group Web Server Web Server Partitioning Data Layer Data Layer
- 12. How to Find Bottleneck
- 13. Cpu load Iowait Log netstat
- 15. Consistency
- 16. Master-Slave
- 17. One Thing
- 18. Need 4 Servers for disater Master DB Slave Slave Slave
- 19. Need 4 Servers for diaster Master DB Service Service Fail Slave Slave Slave
- 20. Need 4 Servers for diaster Master DB Service Dump New Slave Slave Slave Copy
- 21. Need 4 Servers for disater Master DB Slave Slave Slave
- 22. Many Person Knows This Model
- 24. Many Read, Small Write, and No Consistency Write Master DB Async Replication Read Read Slave
- 25. Many Read, Small Write, and No Consistency Write ‘Data’ Master DB Before Replication Read’Data’ Slave Read Past Data
- 26. Many Read, Small Write, and No Consistency Write Master DB Read Read Slave
- 27. How to Support Consistency?
- 29. 2 Phase Commit
- 30. Prepare Phase Master DB Write Write Slave
- 31. Prepare Master DB Write Write Slave Commit Master DB Commit Commit Slave
- 33. Semi-Sync: Send ACK after sending log Write OK Master DB Ack Async Replication Slave
- 34. Mysql Cluster
- 35. Old Version < 7.1 Just Use In-Memory
- 36. Over 7.1 Use In-Memory Index And Disk Data
- 37. Thank you!