scwhittle commented on pull request #15300:
URL: https://github.com/apache/beam/pull/15300#issuecomment-908132864


   Actually I see now you are saying that the amount of memory available for 
processing on a machine with a fixed amount of cpu but increased ram will be 
greater.  That is true, however if the cache itself is the cause of OOMs (not 
user processing) it doesn't help.
   
   With the windmill cache, the objects were cached using the serialized size 
as the weight but the cached objects were deserialized (as that added a large 
benefit to the cache). However this meant that the memory used by the cache 
could be much more than the configured limit depending on the objects cached.  
It seems here we might be caching the protos so perhaps the size estimate is 
reliable.  However if we change to caching user objects in the future we might 
need to be more careful.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to