Skip to content

Clustering And Partitioning

Multitable Clustering

Multitable clustering is great for storing tables that are often joined with each other. In a cluster, tables records are stored interleaved with each other.

dept_name building budget
CSE Taylor 100000
PHY Watson 70000
ID name dept_name salary
100 Hector CSE 65000
101 Robin PHY 57000
102 Seth CSE 58000
103 Genny CSE 70000

These two tables will be merged into the cluster

__ __ __ __
CSE Taylor 100000
100 Hector CSE 65000
102 Seth CSE 58000
103 Genny CSE 70000
PHY Watson 70000
101 Robin PHY 57000

This won't work very well when you're querying for only one table, but makes joins faster. This is stored in variable-length records

Partitioning

In partitioning, a single table X is broken into multiple tables {Xi} . Each table Xi can be stored separately. * This reduces cost of some operations like free space management. * It allows different partitions to be stored on different devices, like storing more important partitions on an SSD and less important ones on a magnetic disk.