Jim C. Nasby wrote:
>How much space does that equate to?
>
>
>Have you optimized the queries?
>
>Items that generally have the biggest impact on performance in
>decreasing order:
>1. System architecture
>2. Database design
>3. (for long-running/problem queries) Query plans
>4. Disk I/O
>5. Memory
>6. CPU
>
>So, I'd make sure that the queries have been optimized (and that
>includes tuning postgresql.conf) before assuming you need more hardware.
>
>Based on what you've told us (very little parallelization), then your
>biggest priority is probably either disk IO or memory (or both). Without
>knowing the size of your database/working set it's difficult to provide
>more specific advice.
>
>
Hi!
We have 3 Compaq Proliant ML530 servers with dual Xeon 2.8GHz
processors, 3 GB DDR RAM, Ultra Wide SCSI RAID5 10000rpm and 1000Gbit
ethernet. We partitioned our databases among these machines, but there
are cross refrences among the machines theoretically. Now the size of
datas is about 100-110GB. We've used these servers for 3 years with
Debian Linux. We have already optimized the given queries and the
postgresql.conf. We tried more tricks and ideas and we read and asked on
mailing lists. We cannot do anything, we should buy new server for the
databases, because we develop our system for newer services, so the size
will grow along. After that we need better responsiblility and shorter
execution time for the big queries (These queries are too complicated to
discuss here, and more times we optimized with plpgsql stored procedures.).
The PostgreSQL 8.1 solved more paralellization and overload problem, the
average load is decreased significantly on our servers. But the big
queries aren't fast enough. We think the hardver is the limit. Generally
2 parallel guery running in working hours, after we make backups at night.
Regards, Atesz
