Currently I can get no more than 40k records/second to be copied
in to postgres. This is also the case when running multiple
cocurrent copy in commands from different connections. It
seems linear in that 1 process will copy in at 40k/s and 2
will copy in at about 20k/s, etc.
I need to know whether I can get better performance from
the db with the hardware specified below. Changing from
mirrored to stripe RAID is not an option :( at this time
as is turning off fsync (which did not work much better).
configure --libdir=/usr/lib \
OS: RedHat 9
sdc1: /pg_data (RAID 1)
hdparm output shows average of 45MB/s
iostat shows fairly constant 10MB/s written to pg_data
and 5-8MB/s to WAL.
CPU: 2x 2.0G Xeon
pgsql-admin by date
|Next:||From: Bruno Wolff III||Date: 2004-04-14 14:57:53|
|Subject: Re: Real time replication of PG DBs accross two servers -|
|Previous:||From: Ben Kim||Date: 2004-04-14 14:34:10|
|Subject: Re: sequence value of the record just inserted.|