| From: | Gregory Stark <stark(at)enterprisedb(dot)com> |
|---|---|
| To: | "Simon Riggs" <simon(at)2ndquadrant(dot)com> |
| Cc: | "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us>, "Heikki Linnakangas" <heikki(at)enterprisedb(dot)com>, "Pavan Deolasee" <pavan(dot)deolasee(at)gmail(dot)com>, "PostgreSQL-development" <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: SetBufferCommitInfoNeedsSave and race conditions |
| Date: | 2007-07-05 08:05:18 |
| Message-ID: | 871wfnwa0x.fsf@oxford.xeocode.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
"Simon Riggs" <simon(at)2ndquadrant(dot)com> writes:
> I'd guess that storing 8 per page would be optimal, so each stored xid would
> track 4,000 transactions - probably around 1 sec worth of transactions when
> the feature is used.
This is something we can experiment with but I suspect that while 8 might be
sufficient for many use cases there would be others where more would be
better. The cost to having more lsns stored in the clog would be pretty small.
On TPCC which has longer transactions on moderate hardware we only see order
of 1,000 txn/min. So a setting like 128 which allows a granularity of 256
transactions would be about 15s which is not so much longer than the xmin
horizon of the 90th percentile response time of 2*5s.
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Heikki Linnakangas | 2007-07-05 11:26:45 | Re: todo: Hash index creation |
| Previous Message | Simon Riggs | 2007-07-05 07:46:32 | Re: SetBufferCommitInfoNeedsSave and race conditions |