[
https://issues.apache.org/jira/browse/HIVE-417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12794263#action_12794263
]
Edward Capriolo commented on HIVE-417:
--------------------------------------
I currently am benching an 11 node hive cluster against a 16 TB MySQL system
4x quad core 32 GB RAM 5.1 with partitioning.
Hive destroys mysql with any query like:
(date_id is my partition column.)
{noformat}
set mapred.map.tasks=34;
set mapred.reduce.tasks=11;
FROM pageviews
insert overwrite directory '/user/ecapriolo/hivetest4'
select sitename_id, user_id, count(user_id) WHERE date_id=20091250 group by
sitename_id,user_id
12098855 Rows loaded to /user/ecapriolo/hivetest4
OK
Time taken: 185.528 seconds
{noformat}
The same query can take over 3000 seconds on MySQL because these large summary
queries are always written to a temp table and then writes bottleneck your read
queries.
However, if mysql has an index (and if the index is in memory, which is hard in
a warehouse) on some other value in the where clause like:
{noformat}
select sitename_id, user_id, count(user_id) WHERE date_id=20091250 and
sitename_id=400 group by sitename_id,user_id
{noformat}
MySQL gets a relative performance speed-up, while hive ends up scanning the
entire table.
I agree with dhruba,
>>This sounds really awesome! Make hadoop-hive suitable for things other than
>>brute force table-scans!
If we had indexes helping stop some brute force scans, that would just open up
other doors to what hive could do.
> Implement Indexing in Hive
> --------------------------
>
> Key: HIVE-417
> URL: https://issues.apache.org/jira/browse/HIVE-417
> Project: Hadoop Hive
> Issue Type: New Feature
> Components: Metastore, Query Processor
> Affects Versions: 0.2.0, 0.3.0, 0.3.1, 0.4.0
> Reporter: Prasad Chakka
> Assignee: He Yongqiang
> Attachments: hive-417.proto.patch, hive-417－2009-07-18.patch
>
>
> Implement indexing on Hive so that lookup and range queries are efficient.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
