From: Jamie f. <ja...@in...> - 2004-12-01 18:16:59
|
Thanks for the feedback Alan. Along those lines, I've just done something similar to help with the WAL's on PGSQL. Increased wal_buffers to 64, and put the pg_xlog directory on a separate physical disk (WAL writes separate from the data files), and got 60KB/s improvement on backup rate, which is approx 8% faster. This got virtually the same speed improvement as turning fsync off on an identical backup run (slightly better than disabling fsync entirely), so I can leave fsync on. Jamie > On the subject of PgSQL vs MySQL... > > This data is several years old now, but a colleague and I benchmarked > MySQL vs Postgres while evaluating processes running several million > inserts/reads and massive concurrency. (It was the forerunner of one of > the most popular DNSBLs of the time.) > > The end results: > > MySQL is mostly faster in inserts because Postgres does a fsync() after > every insert - safer, but slower than MySQL, which doesn't. > > MySQL is optimised for webserver use - results are best on repetitive, > simple queries, it falls away rapidly on complex queries, lots of inserts > or lots of parallel accesses - and especially when there are combinations > of these (300+ processes accessing a MySQL server is _ugly_). > > Postgres' fsync() can be disabled and this makes it a _lot_ faster, at > cost of risk of loss of data. > > > I was quite surprised to read the claim that MySQL is optimised for > inserts. This isn't my experience... > > This may all be well out of date. It's been about 4 years since I looked > at the comparisons between these databases, but may serve as a pointer. > > Index creation slows inserts, but makes reads a _lot_ faster. As always > it's a tradeoff about how often you're doing things, but if either > database can now be loaded with inserts which are then committed this > should speed up everything overall. That's the way the Oracle-engined > replacement we were working on was designed before the project was > abandoned. > > Have any database analysts looked at Bacula's DB setup? > > AB > > > > ------------------------------------------------------- > SF email is sponsored by - The IT Product Guide > Read honest & candid reviews on hundreds of IT Products from real users. > Discover which products truly live up to the hype. Start reading now. > http://productguide.itmanagersjournal.com/ > _______________________________________________ > Bacula-users mailing list > Bac...@li... > https://lists.sourceforge.net/lists/listinfo/bacula-users > > !DSPAM:41adf737270699960811368! > > |