Jeffrey created USERGRID-772:
---------------------------------
Summary: Characterize the performance of 2.1
Key: USERGRID-772
URL: https://issues.apache.org/jira/browse/USERGRID-772
Project: Usergrid
Issue Type: Story
Reporter: Jeffrey
Priority: Critical
We need to characterize the performance of 2.1. This will involve using:
- Varying numbers of tomcats
-- we can start with 20 and progressively add more into a load balancer to see
the performance impact
- Varying ES cluster size - starting with 6 and going up to 20
-- we can start with 20 nodes and only have the indexes targeted to a subset
- Varying ES bucket/shard counts - starting with 10
- Varying Cassandra sizes - starting with 6 and going up to 12
General approach should be:
1) Start with 2 tomcats, 6 ES and 6 C*
2) Add +2 Tomcats iteratively until the performance starts to not increase
(implying that either ES or C* is the limiting factor)
3) Add +3 ES nodes and test again. Document impact.
4) Add +3 C* nodes, remove -3 ES nodes and test again. Document impact.
5) Add +3 C* nodes and +3 ES nodes and test again. Document impact.
6) Add +2 Tomcats and see the slope of the TPS graph
7) Add +3 C* - (now at 12) and test again
8) Test iteratively with +2 Tomcats until performance plateaus, up to 20 Tomcats
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
