> By the way > How about distributing OIDs so that the GetNewOidWithIndex function doesn't take long to find an available OID? > For example, skip one and assign an OID. > > As a side effect of this method, the OID progresses faster. > (Maybe the btree index will be fragmented faster) I think the idea of assigning OIDs to sparse is interesting. However, I think it needs to find out how much it will affect the performance. It could be a non-negligible overhead, especially for large amounts of data insertion/updating.
I agree with you. I can't predict the impact of this method on performance. I may need to do some performance tests to find out the impact of this method.
Thank you for your reply.
regards. -- Tomohiro Hiramitsu NTT Open Source Software Center