Sharding
Download as KEY, PDF1 like789 views
Sharding is a technique for scaling databases by partitioning data across multiple database servers or nodes. There are different ways to implement sharding such as sharding on the primary key or an index in a relational database. For key-value stores, a common approach is to hash the key and assign it to a node using consistent hashing. While sharding improves scalability, it also introduces some limitations like not being able to perform joins across shards and additional work required for data maintenance.
![Key / Value
• server = serverlist[ hash(key) %
len(serverlist) ]](https://image.slidesharecdn.com/sharding-120928021008-phpapp01/85/Sharding-4-320.jpg)
More Related Content
Sharding
- 2. • Sharding is about scaling • scaling up VS scaling out
- 3. Scaling Out • Add more nodes • Each node shared nothing
- 4. Key / Value • server = serverlist[ hash(key) % len(serverlist) ]
- 5. Relational DB • On Primary Key • On Index
- 6. Middle layer • MySQL Proxy
- 7. Client Side • Memcached client • ORM
- 8. Limitation • No joining • Join on the same shard is OK • Data maintenance
- 9. Pre-Sharding • Move database is easier than move data
- 10. Problem of sharding • serverlist[ hash(key) % len(serverlist)
- 11. Consistent Hash • ketama from last.fm • De facto Standard for memcached
- 12. Do we use sharding? • Memcached Client • MySQL for different ventures
- 13. SQL vs NoSQL • NoSQL is “Not Only SQL” • NoSQL completes SQL, but not replaces it
- 14. NoSQL is about • Performance? • Scaling (easier to shard)? • Flexibility (schema-less)?
- 15. Performance • Memcacahed is faster than MySQL? • SQL faster than K/V? • Memory faster than disk?
- 16. MySQL for K/V • handler socket
- 17. Schema - less • Friendfeed’s solution • Table as secondary index, then shard
- 18. • No silver bullet • Trade off