Limitations of tiny partitions

I am having around 150 million rows in a table. Each row is a partition which means 150 million partitions. Size of each partition is 24kb.

I know that small partitions work perfectly for even distribution of data without hotspots. But, having partitions as small as 24kb can be a problem in managing them in cache (c++ objects limitations) or using mv’s/secondary index for this table in scylla.

Is this true? Is this improved in latest version of scylla?

No problem. ScyllaDB is designed to support billions of partitions.

Thanks for your reply.

Just to cross check, tiny partitions won’t be blocker for

  • materialized view’s and secondary indexes during repairs.
  • caching tiny partitions in memory

Thanks.

You should be fine. Plus, if needed, you can always tune bloom filter, repair intensities, and compaction settings in case you find anything weird later on.

1 Like