Subject | Re: [Firebird-Architect] Can we, can we, can we????... |
---|---|
Author | Jim Starkey |
Post date | 2005-06-14T11:28:38Z |
Dmitry Yemanov wrote:
in a loaded system since it depends on things you can't predict or even
observe. But you say you can't predict the number of fetches, which is
regularly reported as part of user feedback and a direct function of the
work performed by the request.
OK, a compromise. We use "execution units", an arbitrary measure of
work performed. An execution unit approximately a quarter of the work
required to fetch a record.
>"Jim Starkey" <jas@...> wrote:That makes no sense to me. You can't possible predict the elapsed time
>
>
>>Is a timeout elapsed time or cpu time?
>>
>>
>
>The former.
>
>
>
>>How about limiting the number of fetches?
>>
>>
>
>Nope, this is not something one could estimate.
>
>
>
>
in a loaded system since it depends on things you can't predict or even
observe. But you say you can't predict the number of fetches, which is
regularly reported as part of user feedback and a direct function of the
work performed by the request.
OK, a compromise. We use "execution units", an arbitrary measure of
work performed. An execution unit approximately a quarter of the work
required to fetch a record.