Обсуждение: Any good olap benchmarks?
Hi! I'm searching for any good OLAP-like benchmark: I need some benchmark with complex selects run on 10-30G dataset, something like this: http://www.percona.com/docs/wiki/benchmark:wikistat:start, but this is only a draft and would only be released for mysql. I've found DBT-3 benchmark, but it lacks documentation and I have no idea how to run in over network (I need server and client to be on different hosts). -- Vladimir Rusinov Google Profile: http://www.google.com/profiles/vladimir.rusinov
Vladimir Rusinov wrote: > I'm searching for any good OLAP-like benchmark: I need some benchmark > with complex selects run on 10-30G dataset, something like this: > http://www.percona.com/docs/wiki/benchmark:wikistat:start, but this is > only a draft and would only be released for mysql. > I already intended to convert and run Vadim's Wikipedia statistics benchmark when I get to it; as we're busy getting the next PostgreSQL release out the door right now I just haven't gotten to it yet. I expect we can get that added into their mix on the same hardware once I'm done. I've linked to everything I'm aware of at http://wiki.postgresql.org/wiki/Category:Benchmarking and the TPC-H page has probably the most relevant information for OLAP. PostgreSQL doesn't do particularly well on OLAP benchmarks yet due to how queries are limited by a single processor, making it hard to get excited about publishing the results. Greenplum ran some interesting tests of their own software against PostgreSQL at http://community.greenplum.com/showthread.php?t=113 you might find interesting. That includes a fairly easy to use TPC-H like test kit program, and by showing where they did much better is suggest the areas that community Postgres struggles relative to software that handles parallel query across multiple cores/servers. -- Greg Smith 2ndQuadrant US Baltimore, MD PostgreSQL Training, Services and Support greg@2ndQuadrant.com www.2ndQuadrant.us
Zitat von Greg Smith <greg@2ndquadrant.com>: > I've linked to everything I'm aware of at > http://wiki.postgresql.org/wiki/Category:Benchmarking and the TPC-H > page has probably the most relevant information for OLAP. > PostgreSQL doesn't do particularly well on OLAP benchmarks yet due > to how queries are limited by a single processor, making it hard to > get excited about publishing the results. > > Greenplum ran some interesting tests of their own software against > PostgreSQL at http://community.greenplum.com/showthread.php?t=113 > you might find interesting. That includes a fairly easy to use > TPC-H like test kit program, and by showing where they did much > better is suggest the areas that community Postgres struggles > relative to software that handles parallel query across multiple > cores/servers. I maybe wrong but this looks like "partitioning" PostgreSQL to me???
I am looking ahead to when postgresql 9.x is non-beta and how we can best make use of the "virtually synchronous" replication. At our site, we have a load balancer appliance that works very well in that it can direct a connection to one of a number of back ends. When one back end goes away, connections will be created to the remaining back ends. I think that this will work well for the postgres failover, however, the database connection will drop (I don't see how existing connections can be failed to the hot-standby postgres server) and need to be established again. Since I can not rely on the applications to perform the reconnect, I was wondering if anyone knows if PgPool-II can accomplish this for me. If it can, then the applications will not be aware that they have been reconnected, they will just be aware that any outstanding transaction failed. Can anyone comment on this functionality of PgPool-II or on this approach to failover in general? Evan Rempel.