Subject | Re: [Firebird-Architect] Group Commits |
---|---|
Author | Pascalis |
Post date | 2004-11-08T13:02:07Z |
Hi Jim,
I agree with you that at most cases it does not make any sense but there are
exceptions. I can give you some examples I can thing of now:
1) There are applications where a unique sequential number (without holes)
is required on every transaction. In this case as you well understand every
transaction waits previous one to commit.
2) When inserting large amount of data I usually commit every few hundreds
of records. This has to be rewritten.
I agree with you that at most cases it does not make any sense but there are
exceptions. I can give you some examples I can thing of now:
1) There are applications where a unique sequential number (without holes)
is required on every transaction. In this case as you well understand every
transaction waits previous one to commit.
2) When inserting large amount of data I usually commit every few hundreds
of records. This has to be rewritten.
----- Original Message -----
From: "Jim Starkey" <jas@...>
To: <Firebird-Architect@yahoogroups.com>
Sent: Monday, November 08, 2004 2:32 PM
Subject: Re: [Firebird-Architect] Group Commits
Pascalis wrote:
>This also means that any single client application cannot make more than 5
>commits/sec. It may be unacceptable on some cases.
>
>
>
What cases and why? I'm not being cynical, I just want to understand
the circumstances under which this would be important?
There are clearly cases where the wait doesn't make sense. If the
transaction committing is the only active transaction, the probability
that another will start and finish within the interval is slight. There
may be other cases we could identify.
Yahoo! Groups Links