Re: Evaluating overall performance
From: Shakespeare <whatsin_at_xs4all.nl>
Date: Thu, 08 Apr 2010 22:49:02 +0200
Message-ID: <4bbe4145$0$22917$e4fe514c_at_news.xs4all.nl>
Op 8-4-2010 19:28, Mladen Gogala schreef:
> On Thu, 08 Apr 2010 07:58:09 -0700, AD wrote:
>
>> ..sort of the opposite of "tuning" for a specific tasks, and more like
>> assessing the actual performance of the server...
>
> Geee, I am trying to fight this myth for years and it still persists.
> There is nothing even resembling "the actual performance". You can either
> be in one of 2 states: users are satisfied with the application
> performance or users are dissatisfied with the application performance.
> You can only measure the application performance, not the "actual
> performance of the server", "overall performance of the server" or "pink
> elephants of the server". My favorite method of estimating the overall
> performance of the server involves the gentleman named Sam Adams, in pint
> sized dozes. Please, remember, no database has ever been problematic
> without some help from users and data. No users and no data equals no
> problems. You cannot measure the performance of a system, database or
> instance. About the only thing you can and do measure is the application
> performance. So called "big picture" becomes apparent only after
> sufficient socializing with Mr. Adams. It usually causes rather nasty
> hangover. I prefer concentrating on the applications. That's my job: to
> ensure the availability of the database and the optimal performance of
> the applications accessing my database.
>
>
>
Mladen,
Date: Thu, 08 Apr 2010 22:49:02 +0200
Message-ID: <4bbe4145$0$22917$e4fe514c_at_news.xs4all.nl>
Op 8-4-2010 19:28, Mladen Gogala schreef:
> On Thu, 08 Apr 2010 07:58:09 -0700, AD wrote:
>
>> ..sort of the opposite of "tuning" for a specific tasks, and more like
>> assessing the actual performance of the server...
>
> Geee, I am trying to fight this myth for years and it still persists.
> There is nothing even resembling "the actual performance". You can either
> be in one of 2 states: users are satisfied with the application
> performance or users are dissatisfied with the application performance.
> You can only measure the application performance, not the "actual
> performance of the server", "overall performance of the server" or "pink
> elephants of the server". My favorite method of estimating the overall
> performance of the server involves the gentleman named Sam Adams, in pint
> sized dozes. Please, remember, no database has ever been problematic
> without some help from users and data. No users and no data equals no
> problems. You cannot measure the performance of a system, database or
> instance. About the only thing you can and do measure is the application
> performance. So called "big picture" becomes apparent only after
> sufficient socializing with Mr. Adams. It usually causes rather nasty
> hangover. I prefer concentrating on the applications. That's my job: to
> ensure the availability of the database and the optimal performance of
> the applications accessing my database.
>
>
>
Mladen,
for a (most) part you're right. But instead of waiting for users to become dissatisfied, some characteristics of a server may predict this before they start complaining. I don't know mister Adams, but guess it's some kind of booze, right?
Shakespeare Received on Thu Apr 08 2010 - 15:49:02 CDT