We have modified our field creation mechanism from not analyzed field to multi valued field (one analyzed and one not analyzed).
After changing this mechanism we are now seeing increase in disk space (around 15%) and heap memory (around 35%).
Disk space storage increase is inline with our expected analysis. But this much of heap memory increase was not expected.
After drilling down usage of heap memory we found that consumption was mainly increase due to more memory usage by 'BlockTreeTermsReader'
- Is this heap memory increase is due to usage of multi valued field? (From analysis it seems because of it only, just want to confirm)
- Is there any optimization we can do to avoid it ?