Re: [PERFORM] postgres performance tunning
My requirement is more than 15 thousand queries will run, It will be 5000 updates and 5000 insert and rest will be select. Each query will be executed in each psql client, (let say for 15000 queries 15000 thousand psql connections will be made). Since the connections are more for me the performance is low, I have tested it with pgbench tool. Configurations, RAM : 17.9GB CPU : 64-bit 2 cores each 5346 bogomips Postgres Configurations, Shared Memory Required (shmmax) : 172032 bytes Wal Buffers : 1024KB Maintenance work mem : 1024MB Effective Cache Size : 9216MB Work Memory : 32MB Shared Buffer : 1536MB -- View this message in context: http://postgresql.1045698.n5.nabble.com/postgres-performance-tunning-tp3307846p3309251.html Sent from the PostgreSQL - performance mailing list archive at Nabble.com. -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Compared MS SQL 2000 to Postgresql 9.0 on Windows
If you strictly have an OLTP workload, with lots of simultaneous connections issuing queries across small chunks of data, then PostgreSQL would be a good match for SQL server. This matches my observations. In fact, PostgreSQL's MVCC seems to work heavily in my favor in OLTP workloads. On the other-hand, if some of your work load is OLAP with a few connections issuing complicated queries across large chunks of data, then PostgreSQL will not perform as well as SQL server. SQL server can divide processing load of complicated queries across several processor, while PostgreSQL cannot. While I agree with this in theory, it may or may not have a big impact in practice. If you're not seeing multi-cpu activity spike up on your MSSQL box during complex queries, you aren't likely to benefit much. You can test by timing a query with and without a query hint of MAXDOP 1 select * from foo with (MAXDOP = 1) which limits it to one processor. If it runs just as fast on one processor, then this feature isn't something you'll miss. Another set of features that could swing performance in MSSQL's favor are covering indexes and clustered indexes. You can sort-of get around clustered indexes being unavailable in PostgreSQL - especially on low-churn tables, by scheduling CLUSTER commands. I've seen discussions recently that one or both of these features are being looked at pretty closely for inclusion in PostgreSQL. -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] postgres performance tunning
Thanks for ur suggestion, already I have gone through that url, with that help I was able to make my configuration to work for 5K queries/second. The parameters I changed was shared_buffer, work_mem, maintenance_work_mem and effective_cache. Still I was not able to reach my target. Can u kindly tell me ur postgres configurations thereby I can get some idea out of it. -- View this message in context: http://postgresql.1045698.n5.nabble.com/postgres-performance-tunning-tp3307846p3310337.html Sent from the PostgreSQL - performance mailing list archive at Nabble.com. -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Strange optimization - xmin,xmax compression :)
On Dec 17, 2010, at 8:46 PM, Robert Haas wrote: 2010/12/6 pasman pasmański pasma...@gmail.com: hello. i tested how are distributed values xmin,xmax on pages. in my tables . typically there are no more than 80 records on pages. maybe its possible to compress xmin xmax values to 1 byte/per record (+table of transactions per page)? its reduce the space when more than 1 record is from the same transaction. Not a bad idea, but not easy to implement, I think. Another option that would help even more for data warehousing would be storing the XIDs at the table level, because you'll typically have a very limited number of transactions per table. But as Robert mentioned, this is not easy to implement. The community would probably need to see some pretty compelling performance numbers to even consider it. -- Jim C. Nasby, Database Architect j...@nasby.net 512.569.9461 (cell) http://jim.nasby.net -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] CPU bound
On 12/19/2010 7:57 PM, James Cloos wrote: RA == Royce Ausburnro...@inomial.com writes: RA I notice that when restoring a DB on a laptop with an SDD, RA typically postgres is maxing out a CPU - even during a COPY. The time the CPUs spend waiting on system RAM shows up as CPU time, not as Wait time. It could be just that the SSD is fast enough that the RAM is now the bottleneck, although parsing and text=binary conversions (especially for integers, reals and anything stored as an integer) also can be CPU-intensive. -JimC Good time accounting is the most compelling reason for having a wait event interface, like Oracle. Without the wait event interface, one cannot really tell where the time is spent, at least not without profiling the database code, which is not an option for a production database. -- Mladen Gogala Sr. Oracle DBA 1500 Broadway New York, NY 10036 (212) 329-5251 www.vmsinfo.com -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance