You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "Karen Smoler Miller (JIRA)" <ji...@apache.org> on 2017/09/25 22:48:00 UTC

[jira] [Created] (GEODE-3701) Improve docs on hash index performance

Karen Smoler Miller created GEODE-3701:
------------------------------------------

             Summary: Improve docs on hash index performance
                 Key: GEODE-3701
                 URL: https://issues.apache.org/jira/browse/GEODE-3701
             Project: Geode
          Issue Type: Improvement
            Reporter: Karen Smoler Miller


Stronger language is warranted for hash indexes.
In a simple test, there was an order of magnitude difference in load and GII performance between the two types of indexes.

Range:
Client: Loaded 2,000,000 entries in 20,487 ms
Server 1 Size: 1,272,322,680 bytes
Server 2: GIIed in 12,523 ms
Server 2 Size: 1,272,088,544 bytes

Hash:
Client: Loaded 2,000,000 entries in 257,016 ms
Server 1 Size: 1,228,010,280 bytes
Server 2: GIIed in 218,983 ms
Server 2 Size: 1,228,161,664 bytes
So, the difference in size was ~45MB, but the difference in time was an order of magnitude.

Better wording suggested for file developing/query_index/creating_hash_indexes.html:
Using hash index will degrade put performance and recovery time substantially. If memory is not a concern, it is recommended that you use range index.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)