[
https://issues.apache.org/jira/browse/CASSANDRA-20942?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stefan Miklosovic updated CASSANDRA-20942:
------------------------------------------
Description:
Extend the traincompressiondictionary nodetool command to include global /
local dictionary scope
* Global (default): Store in system_distributed.compression_dictionaries and
make available cluster-wide.
* Local: Restrict dictionary availability to the current node, when dictionary
from local data is desired.
This should be safe in general as dictionaries are baked into an SSTable and
"travel" (e.g. by means of streaming) around cluster. The fact another node
does not have a dictionary of another to compress is okay, if it ever gets an
SSTable from another node, it will be able to work with it anyway.
This option might be choosen for the evaluation of a dictionary compression
ratio before rolling it out cluster-wide. That would be done by e.g.
"exporting" and "importing" via nodetool which will save it to
system_distributed ...
was:
Extend the traincompressiondictionary nodetool command to include:
* Support for multiple sampling strategies.
* '--global' and '--local' options to define dictionary availability scope:
** Global (default): Store in system_distributed.compression_dictionaries and
make available cluster-wide.
** Local: Restrict dictionary availability to the current node, when
dictionary from local data is desired.
> Enhance traincompressiondictionary command with sampling and scope Options
> --------------------------------------------------------------------------
>
> Key: CASSANDRA-20942
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20942
> Project: Apache Cassandra
> Issue Type: Improvement
> Reporter: Yifan Cai
> Assignee: Yifan Cai
> Priority: Normal
>
> Extend the traincompressiondictionary nodetool command to include global /
> local dictionary scope
> * Global (default): Store in system_distributed.compression_dictionaries and
> make available cluster-wide.
> * Local: Restrict dictionary availability to the current node, when
> dictionary from local data is desired.
> This should be safe in general as dictionaries are baked into an SSTable and
> "travel" (e.g. by means of streaming) around cluster. The fact another node
> does not have a dictionary of another to compress is okay, if it ever gets an
> SSTable from another node, it will be able to work with it anyway.
> This option might be choosen for the evaluation of a dictionary compression
> ratio before rolling it out cluster-wide. That would be done by e.g.
> "exporting" and "importing" via nodetool which will save it to
> system_distributed ...
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
