On Mon, Feb 22, 2010 at 1:13 PM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> wrote:
> hello
>
> the speed depends on setting of working_memory. Try to increase a working_memory
>
> set working_memory to '10MB';
It's already at
tpcc=# show work_mem;
work_mem
----------
20000kB
(1 row)
I also wouldn't have imagined an external merge-sort as being very
memory-intensive--wouldn't it only enough buffer space to read 2x and
write 1x in big-enough chunks for mostly-sequential access?
--
Yang Zhang
http://www.mit.edu/~y_z/