[
https://issues.apache.org/jira/browse/KYLIN-2867?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhong Yanghong updated KYLIN-2867:
----------------------------------
Description:
Too many fuzzy keys in one scan may downgrade the performance. Currently if the
number of fuzzy keys exceeds a threshold, like 200, the scan will not use fuzzy
key. However, this may also downgrade the performance.
There are many cases in eBay that the total number of fuzzy keys is around
1000. To deal with these cases, it's better to split the large fuzzy key set,
like 5*200. Then for each subset, assign a scan. This strategy will increase
the number of scan to reduce the number of fuzzy keys in each scan to improve
the total performance.
> split large fuzzy Key set
> -------------------------
>
> Key: KYLIN-2867
> URL: https://issues.apache.org/jira/browse/KYLIN-2867
> Project: Kylin
> Issue Type: Improvement
> Components: Query Engine
> Affects Versions: v2.1.0
> Reporter: Zhong Yanghong
> Assignee: Zhong Yanghong
>
> Too many fuzzy keys in one scan may downgrade the performance. Currently if
> the number of fuzzy keys exceeds a threshold, like 200, the scan will not use
> fuzzy key. However, this may also downgrade the performance.
> There are many cases in eBay that the total number of fuzzy keys is around
> 1000. To deal with these cases, it's better to split the large fuzzy key set,
> like 5*200. Then for each subset, assign a scan. This strategy will increase
> the number of scan to reduce the number of fuzzy keys in each scan to improve
> the total performance.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
