xiaojian zhou created GEODE-6973:
------------------------------------
Summary: getExistingIdForType should not compare all entries in
idToType region
Key: GEODE-6973
URL: https://issues.apache.org/jira/browse/GEODE-6973
Project: Geode
Issue Type: Bug
Reporter: xiaojian zhou
We found the PeerTypeRegistration's getExistingIdForType() will iterate through
the idToType region's entries to find if the incoming newType is there.
If idToType region contains 20K or 100K entries, this will impact the put
throughput (customers did notice the performance downgrade when there're many
pdxTypes).
To make the things worse, the comparison is to compare the whole object, field
to field. If the json object (which will be converted to pdxType) contains 30
fields, the comparison will have to compare up to 30 fields. If the idToType
region contains 20K entries, A new pdxType will do 20K x 30 string comparisons
before register it.
To speed up, we proposed to add an index hashmap, which used pdxType's hashcode
as key, pdxTypeId's arrayList or hashset as value.
When a new pdxType arrived, calculate its hashcode, then use the hashcode to
find its pdxTypeIds (it could be a list if there're hash conflicts, but in most
of case, it should be only one). Then use this small list to find the pdxTypes.
Only need to do field-to-filed compare to this small list of pdxType.
This algorithm will dramatically speed up.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
