Details
-
Improvement
-
Resolution: Duplicate
-
Major
-
None
-
4.5.1
Description
The difference between the high watermark and low watermark is pretty high (10%) and the cost of pulling items back off disk is costly. In a few production environments where there is slow growth, when the cluster hits the high watermark it has a massive impact on the workload. I suspect it would be better if the low watermark is much closer to the high. Scanning memory a few more time will always be faster than pulling data from disk.
Attachments
Issue Links
- duplicates
-
MB-4249 Better calculate high/low watermarks for large RAM systems
- Reopened