[ 
https://issues.apache.org/jira/browse/PHOENIX-933?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14061087#comment-14061087
 ] 

ASF GitHub Bot commented on PHOENIX-933:
----------------------------------------

Github user chrajeshbabu commented on a diff in the pull request:

    https://github.com/apache/phoenix/pull/3#discussion_r14898287
  
    --- Diff: 
phoenix-core/src/main/java/org/apache/phoenix/iterate/DefaultParallelIteratorRegionSplitter.java
 ---
    @@ -140,7 +142,14 @@ public boolean apply(HRegionLocation location) {
             // distributed across regions, using this scheme compensates for 
regions that
             // have more rows than others, by applying tighter splits and 
therefore spawning
             // off more scans over the overloaded regions.
    -        int splitsPerRegion = regions.size() >= targetConcurrency ? 1 : 
(regions.size() > targetConcurrency / 2 ? maxConcurrency : targetConcurrency) / 
regions.size();
    +        PTable table = tableRef.getTable();
    --- End diff --
    
    bq. why do we need this to be different for local indexes? Seems like we 
could run parallel scans over part of each region just like we do with other 
scans, no?
    Index table rows have region start key as prefix so even if we split region 
key ranges into multiple and scan parallelly,only first scanner gives the 
results and remaining all the scanners just return nothing. So number of splits 
of local index region setting to 1.
    if we want to start multiple scanners for local index into multiple parts 
then we need to split the scan ranges into multiple.  


> Local index support to Phoenix
> ------------------------------
>
>                 Key: PHOENIX-933
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-933
>             Project: Phoenix
>          Issue Type: New Feature
>            Reporter: rajeshbabu
>
> Hindex(https://github.com/Huawei-Hadoop/hindex) provides local indexing 
> support to HBase. It stores region level index in a separate table, and 
> co-locates the user and index table regions with a custom load balancer.
> See http://goo.gl/phkhwC and http://goo.gl/EswlxC for more information. 
> This JIRA addresses the local indexing solution integration to phoenix.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to