> .. I think Maxim might need to try tweaking effective_cache_size. ..
I'm right now doing some troubleshooting for Bitmap Index Scan vs Index
Scan and effective_cache_size caught my attention. I want to take this
chance to confirm my understanding about this parameter.
1. By design, effective_cache_size should not less than shared_buffer
since file-system cache should be considered as well.
2. Mackert and Lohman's algorithm in index_pages_fetched looks doesn't
consider the number of concurrency? so if we have N sessions to access N
different small relations, but sum_of_n_relations is larger than
effective_cache_size, would the index_pages_fetched figure out a much
smaller value than the fact?
--
Best Regards
Andy Fan