Jacek Lewandowski created CASSANDRA-18201:
---------------------------------------------
Summary: Store min and max partition key in sstable stats metadata
rather than in some index component
Key: CASSANDRA-18201
URL: https://issues.apache.org/jira/browse/CASSANDRA-18201
Project: Cassandra
Issue Type: Improvement
Components: Local/SSTable
Reporter: Jacek Lewandowski
Assignee: Jacek Lewandowski
Currently min and max partition key is stored in the index summary.
Firstly, that informat better fits stats metadata as there are other similar
statistics (like min and max clusterings)
Secondly, opening index summary is costly. Though, index summary and the index
itself are loaded upon opening an sstable just because we need to read min and
max partition keys. Min and max partition keys need to be know for an sstable
so that when some data are queried, we can select the sstable which may contain
that data - that is, whether the queried partition is included in the min/max
key range of the sstable.
With the proposed solution, we could postpone loading index components to the
time when the data from such sstable is really requested. It will be enough to
read lightweight stats metadata as it will be sufficient to know everything
about data ranges included in that sstable. It will also let to save memory
used by those components until data from those sstables are requested.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
