[SysD] Data Partition/Sharding, CAP theorem

Why?

Database partitioning is normally done for manageability, performance or availability^[1] reasons, or for load balancing.

How?

• Range partitioning example: from 0 - 100, then 101 - 200....
• List partitioning ex: user defined list into one partition : all europ countries
• Composite partitioning: or example first applying a range partitioning and then a hash partitioning
• Round-robin partitioning : by order : 0 to p0, 1 to p1, 2 to p2 ...
• Hash partitioning: hash and modulo

Tow directions:

• Horional partition:
• partition of rows with a key
• Vertical partition
• column spliting
• "normalization"
•  A columnar database can be regarded as a database that has been vertically partitioned until each column is stored in its own table.

Problemes:

1. rebalancing data : a shard is outgrown and we need to split it 
2. Joining data from multiple shards
3. how to partition data in a shard

在理論計算機科學中，CAP定理（CAP theorem），又被稱作布魯爾定理（Brewer's theorem），它指出對於一個分布式计算系統來說，不可能同時滿足以下三點：^[1][2]

• 一致性（Consistency） （等同于所有节点访问同一份最新的数据副本）
• 可用性（Availability）（每次请求都能获取到非错的响应——但是不保证获取的数据为最新数据）
• 分区容错性（Partition tolerance）（以实际效果而言，分区相当于对通信的时限要求。系统如果不能在时限内达成数据一致性，就意味着发生了分区的情况，必须就当前操作在C和A之间做出选择^[3]。）

ref:

sharding:  http://highscalability.com/blog/2009/8/6/an-unorthodox-approach-to-database-design-the-coming-of-the.html

https://iswade.github.io/articles/partition/

https://en.wikipedia.org/wiki/Partition_(database)