You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ignite.apache.org by "Vladimir Ozerov (JIRA)" <ji...@apache.org> on 2015/09/11 16:40:45 UTC
[jira] [Created] (IGNITE-1432) .Net: Fix InteropCacheEntryProcessor
performance on remote nodes
Vladimir Ozerov created IGNITE-1432:
---------------------------------------
Summary: .Net: Fix InteropCacheEntryProcessor performance on remote nodes
Key: IGNITE-1432
URL: https://issues.apache.org/jira/browse/IGNITE-1432
Project: Ignite
Issue Type: Task
Components: interop
Affects Versions: 1.1.4
Reporter: Vladimir Ozerov
Priority: Critical
Fix For: ignite-1.5
Considerations:
1) Invoke with single key is expected to be called only once, so no changes is needed here - deploy and execute in a single JNI call.
2) If there are several keys, there is a high chance (but still not 100% due to partitioning) that processor will be called multiple times.
Proposed solution:
1) Check amout of keys.
2) If cnt == 1, no changes to current logic.
3) If cnt > 1, first deploy (JNI call), then execute (JNI call). Processor entry must be put into weak-map located somewhere inside the interop processor. Interop processor must constantly listen for corresponding reference queue and release .Net entries as soon as processor is weakly reacheable.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)