>For cluster-wide setting you will probably want to drop that significantly -- start lower, as in somewhere around 10MB >and work up from there as necessary. For the queries you are finding slow (the reason for these emails) you can set >work_mem specifically for the session.
>Eg.
>set work_mem to '400MB';
><run your query>
>reset work_mem;
I changed it to 10MB.
The problamatic commands runs now less than 10 seconds.
