[
https://issues.apache.org/jira/browse/USERGRID-752?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jeffrey updated USERGRID-752:
------------------------------
Sprint: Usergrid 23, Usergrid 24, Usergrid 25, Usergrid 26 (was: Usergrid
23, Usergrid 24, Usergrid 25, Usergrid 26, Usergrid 28)
> Update Usergrid 1.0 search to use concurrent search per shard
> -------------------------------------------------------------
>
> Key: USERGRID-752
> URL: https://issues.apache.org/jira/browse/USERGRID-752
> Project: Usergrid
> Issue Type: Story
> Reporter: Todd Nine
> Assignee: Todd Nine
> Fix For: 1.0.3
>
>
> Currently we merge all shards together with every page seek in usergrid 1.0.
> This means we select 20k columns, and discard 19k with each page selection.
> Instead, I propose that we create an AST per shard, and execute the iterator
> on these shards concurrently. We then merge on the last step, greatly
> reducing the amount of network I/O and discarded data on queries of large
> data sets.
> Proposed Changes:
> 1 AST instance per shard, evaluated concurrently via some worker pool.
> Change final gather of concurrent iterators.
> Change cursor generation to create the cursor from the merged trees.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
