5

Can anyone please describe the differences between Zipfian and Uniform distribution while running YCSB workloads ?

Here's the YCSB core properties: https://github.com/brianfrankcooper/YCSB/wiki/Core-Properties

Community
  • 1
  • 1
Deepjyoti Saha
  • 63
  • 1
  • 6

1 Answers1

9

The Yahoo Team has explained it in their paper.

In a nutshell, the distribution affects how YCSB reads and scans over the keyspace:

  • uniform: each row has an equal probability to be read
  • zipfian: some rows have more probability to be targeted by reads or scans. Those rows are called "hot set" or "hot spot" and represent popular data, for instance popular threads of a forum. You should set it up with: hotspotdatafraction and hotspotopnfraction. See $YCSB_HOME/workloads/workload_template for more details.

Hope this helps.

Phenyl
  • 495
  • 1
  • 9
  • 13