[prev] 61 [next]

Data Storage in PDBs (cont)

Hash partitioning
  • use hash value to determine which node and page
  • e.g. i = hash(tuple) so tuple is placed on i th node
  • helpful for equality-based queries on hashing attribute
Range partitioning
  • ranges of attr values are assigned to processors
  • e.g. values 1-10 on node0,  11-20 on node1, ...,  99-100 noden-1
  • potentially helpful for range-based queries
In both cases, data skew may lead to unbalanced load