Replies: 1 comment
-
Hello, some time ago index processing was changed to make index maps partitioned (as opposed to global on each node) to save memory and allow parallel query processing on a per-partition basis. This is not in 14.1.1 or 12.2.1.4. That being said, if you think (or measured that) processing is not faster now, we'll be happy to review your app with you. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi
I am involved in an upgrading task where we try to upgrade from Coherence CE 14.1.1.0.18 to Coherence CE 24.09.2. When running a mixed load test which contains both reads and writes I see significantly higher cpu (and slower responses) when running with v24 compared to with v14. In both these setups we use Java v17.0.11 (Corretto) running on RedHat Enterprise Linux v7.9. When cpu profiling, I can see that com.tangosol.util.filter.LikeFilter.applyIndex() is taking > 75% of the cpu time for v24 but a much smaller part for v14. It looks like LikeFilter is calling/triggering operations on PartionedIndexMap for v24 but this not the case when profiling for v14. Is this normal behaviour for v24? Are we missing some migration work where we should modify our java code to work better with v24? The profiling snapshot is taken after ~20 mins of testing so the system should be warmed up.
Beta Was this translation helpful? Give feedback.
All reactions