tag:blogger.com,1999:blog-20038672.post4714323639082021557..comments2024-03-28T00:58:29.187-04:00Comments on Robert Haas: More On Read ScalingRobert Haashttp://www.blogger.com/profile/08393677427643988650noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-20038672.post-65682089458844993122011-07-26T10:26:29.828-04:002011-07-26T10:26:29.828-04:00Actually, this is the perfect time to be making ch...Actually, this is the perfect time to be making changes to this area. Some of the subtle bits Robert has identified here--like the potential issues with CREATE INDEX CONCURRENTLY and Hot Standby--make tweaking this code a scary thing to do. I feel comfortable about this happening now, with >75% fo the release cycle still left for unanticipated changes to fall out of testing. This is not code I'd want to see show at at the end of a release cycle and ship without a long testing period first.Greg Smithhttps://www.blogger.com/profile/16149991742662655368noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-203694473329543642011-07-26T04:37:46.107-04:002011-07-26T04:37:46.107-04:00I think also new patches need to be reviewed with ...I think also new patches need to be reviewed with performance in mind. Like you mentioned, some locks are used by new features and that makes it harder to change them.Greg Jaskiewiczhttps://www.blogger.com/profile/01582040055358451515noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-14192078513667834102011-07-26T01:14:45.397-04:002011-07-26T01:14:45.397-04:00It's hard to accept that we have to wait for P...It's hard to accept that we have to wait for PG 9.2 for these improvements.Bruce Momjianhttps://www.blogger.com/profile/07875088787463864011noreply@blogger.comtag:blogger.com,1999:blog-20038672.post-65836922139694945892011-07-25T22:51:13.796-04:002011-07-25T22:51:13.796-04:00Hey Robert,
I have almost-daily dbt5 (tpc-e deriv...Hey Robert,<br /><br />I have almost-daily dbt5 (tpc-e derivative) tests against the 9.1 and 9.2 branches limping along for almost a month now:<br /><br />http://207.173.203.223/~markwkm/community6/dbt5/dailyv01/dbt5.png<br /><br />My understanding is that we could get twice the transaction rate, at the current scale factor, if the efficiency of the lock manager is improved.<br /><br />If there any interest for you here, I'll do whatever I can to get you the data you need.<br /><br />Markmarkwkmhttps://www.blogger.com/profile/07612969048647363456noreply@blogger.com