"fburgess@xxxxxxxxxxxxxxx" <fburgess@xxxxxxxxxxxxxxx> wrote: > We have 64GB of Memory on RHEL 6.4 > > shared_buffers = 8GB > work_mem = 64MB > maintenance_work_mem = 1GB > effective_cache_size = 48GB > Do these settings look correct for 9.3.2? Maybe. What is your max_connections setting? I find that a good place to start with work_mem is to ignore the factors you quoted, and to set it somewhere near 25% of machine RAM divided by max_connections. It might be possible to go up from there, but monitor closely for peaks of activity which cause enough memory allocation to flush the OS cache and cause high disk read rates, killing performance until the cache repopulates. The can take a while since the high disk read rates slow queries, causing more of them to compete, leading to higher total work_mem allocations, and thus preventing recovery from the performance degradation. In other words, setting this too high leads to unstable performance. It looks better than a lower setting until too many users hit Enter at about the same time, causing performance to collapse for a while. -- Kevin Grittner EDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
