I think the ratio may be a bit of a red herring.

John is taking an extra 40 seconds for 10,000 updates with triggers - about
4ms per call - on a time of 324 w/o triggers - about 32ms per write.

You're taking an extra 600 or so seconds for 500,000 updates with triggers -
about 1ms per call - on a time of 60 secs w/o - about 0.1ms per write.

So your test was 300 times quicker writing data, but only 4 times quicker
running the trigger.

You could change the ratio to nearer to 1 by lowering the write speed as
well as by upping the trigger speed.  What you may want to check is how long
a real-world write takes - your example is quick here because the file is
cached, but it may be the case that you have more un-cached writes from your
application, and so the trigger overhead is less significant.

However - I do think that 1ms per trigger call seems excessive - we're below
0.1ms on our UD / NT system.

Simon



"Dennis Bartlett" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]
> Ok, so something has to be different...
>      we still think you're smoking something...
>
> my ratio is somewhere around 1 to 10
>                your ratio is 1 to 1.15
>
> with your ratio I'd happily embrace triggers.
>
> Please post (possibly privately?) your uvconfig file, what the specs of
> your machine are (ie no of cpus, memory, whether your files are stored
> in RAM...)
>
> basically, if IBM could get their act together on this, and make our IBM
> p615 work like your Sun whatever, we'd be happy as pigs in top dressing!
-------
u2-users mailing list
[EMAIL PROTECTED]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to