AWS OpenSearch Service recommended JVM heap size is often set to 50% of system RAM. If JVM heap size is limited to 32GB, does that mean there is no point in using node types with more than 64GB RAM?
AWS OpenSearch Service recommended JVM heap size is often set to 50% of system RAM. If JVM heap size is limited to 32GB, does that mean there is no point in using machine types with more than 64GB RAM?
If there is a point in having node types with +64GB (ie. 128GB) why? What is the use of the remaining 32GB for a 64GB RAM/ 96GB for a 128GB RAM after the 32GB has been committed to java heap in a node?
May be we can assume that the operating system and other system tasks uses the remaining memory, but how much of that is really needed by the system tasks or OS?
In a nutshell is there any relevance of having huge memory after 32GB of RAM has been committed to java heap for nodes in an OpenSearch cluster? Please assist if you can. Thanks.
References: https://docs.amazonaws.cn/en_us/opensearch-service/latest/developerguide/auto-tune.html https://aws.amazon.com/blogs/big-data/understanding-the-jvmmemorypressure-metric-changes-in-amazon-opensearch-service/
- 新しい順
- 投票が多い順
- コメントが多い順
in general, after 64GB it is indeed makes more sense to scale out rather that up, however the extra memory is always useful as it will be used for OS cache which in turn will speed up queries if relevant segments are cached by operating system.
Hi,
I'm looking for the same answer, then I found the below post on the knowledge center. I believe this can make things clear.
[Why is the SysMemoryUtilization so high on my Amazon OpenSearch Service cluster?]
https://aws.amazon.com/premiumsupport/knowledge-center/opensearch-high-sysmemoryutilization/
Most of the memory used by OpenSearch Service is for in-memory data structures. OpenSearch Service uses off-heap buffers for efficient and fast access to files. The Java virtual machine (JVM) also requires some memory.
Thanks for the feedback @Alext_T. That makes sense.