Hi Jeremy,

Hmmm ... I guess the meat of my suggestion is that it be made
available on a per-transaction basis because in many applications
some transactions are more "critical" than others.  So even on
systems where there are *some* transactions that need to be
flushed to log immediately we can still get some speedups by
delaying log writes for transactions that do not need this...
transactions that can afford to be lost if the server crashes.

And the sad thing is that though it may be simple to implement
I may not be in a position to do so for testing :-(

Seun Osewa


------- Original message ------- 
From: Jeremy Zawodny <[EMAIL PROTECTED]> 
Date: Tue, 30 Sep 2003 10:17:04 -0700 
Subject: Re: Possible Commit Syntax Change for Improved TPS 

On Tue, Sep 30, 2003 at 01:19:19PM +0100, Oluwaseun Osewa  wrote:
> Hi,
>
> I have been studying the basic limitation that the number of
> committed transactions per second possible in a relational databases
> is limited by the number of writed per second of the underlying hard
> disk, since each transaction requires at least the write-ahead log data
>
> to be flushed to disk on commit, thus limiting the number of durable
> transactions commits possible per second.

InnoDB already provides this flexability.

  innodb_flush_log_at_trx_commit={0,1,2}

As described in the InnoDB and MySQL documentation.

It's server-wide, not per-connection or transaction, but it's there
today.

Jeremy
--
Jeremy D. Zawodny     |  Perl, Web, MySQL, Linux Magazine, Yahoo!
  |  http://jeremy.zawodny.com/

MySQL 4.0.15-Yahoo-SMP: up 16 days, processed 586,074,652 queries (405/sec. avg)

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]

SIZE does matter - The UK's biggest *Free* Web based mail - 10 MB Free
mail.lycos.co.uk


-- 
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]

Reply via email to