This is on a CentOS 6.5 box running PostgreSQL 9.2
On 11/13/14 4:09 PM, CS DBA wrote:
> All;
>
> We have a large db server with 128GB of ram running complex functions.
>
> with the server set to have the following we were seeing a somewhat
> low hit ratio and lots of temp buffers
>
> shared_buffers = 18GB
> work_mem = 75MB
> effective_cache_size = 105GB
> checkpoint_segments = 128
>
>
> when we increased the values to these not only did the hit ratio drop
> but query times are now longer as well:
>
>
> shared_buffers = 28GB
> work_mem = 150MB
> effective_cache_size = 105GB
> checkpoint_segments = 256
>
> This does not seem to make sense to me, anyone have any thoughts on
> why more memory resources would cause worse performance?
>
> Thanks in advance
>