Re: Know 1 database, know them all?

[Mogens N�rgaard]

Well, I already started on the SQL Server paper on the R=S+W thing, but I needed help from guys that knew far more than I, and I didn't get it.

Here's the idea:

The formula can be used on statement, session and system level. So that's  3 x 3 grid where at least the R column and the S column can be filled out (numbers found somewhere in various views) while filling out the W stuff on the three levels is where I (and others, probably) need help.

So if someone or somebodies :) can help, we could create the YAPP paper for SQL Server. At least we know what we're hunting, since R-S=W, so it's all about finding as much W as possible, until it about fits.

Mogens

Broodbakker, Mario wrote:

 

Funny..the last time I spoke with the guys in Redmond, they told me: You don't need that...

And OK, I must admit there are a few undocumented wait thingies (viewable with dbcc perfmon according to my old notes..), but they are too cryptic too understand..

I like your R=S+W by the way..You should write a SQLserver paper on that!

 

Mario

-----Original Message-----
From: Mogens Nørgaard [mailto:[EMAIL PROTECTED]]
Sent: dinsdag 18 februari 2003 21:19
To: Multiple recipients of list ORACLE-L
Subject: Re: Know 1 database, know them all?

Not quite true, as far as I know.... There's wait stuff in there, although not enough to my taste. There's cpu in there, and the start and stop time, which makes it possible to at least make a crude R = S + W, where the difficult part is breaking down the W into meaningful stuff. A long way to go, but I think they're aware of it. The guys from SQL Server Development I spoke to about it were very interested in the method and liked the whole idea.

But don't forget that I'm always wrong.

Mogens

Broodbakker, Mario wrote:

Mogens, the only problem with your statement about 'oracle myths & king of the new world' is that the only way of looking at SQLServer performance is probably looking at ratio's: there are no wait statistics, there is one(1) latch wait counter though! for the complete system :( (apart from some other almost useless perfmon counters, taht is..)

regards,
Mario
 

-----Original Message-----
Sent: maandag 17 februari 2003 23:29
To: Multiple recipients of list ORACLE-L


I see it from a slightly different (and probably wrong) angel, at least 
regarding the performance of things and databases: If you've worked with 
Oracle databases for some time (and have real experience), and know 
about the myths and their anti-thesis (use the wait interface instead of 
the ¤&#% ratio crap, know about RAID-5, don't have too many indexes, 
concentrate on LIO instead of PIO, etc.,etc.) you'll do quite fine. As 
Peter Gram once said to me: It's all about getting a database to perform 
on a platform.

You can take your old presentations regarding Oracle myths and change it 
into a SQL Server or mySQL presentation, change a few details, and be 
king in the new world.

Mogens

Robert Eskridge wrote:

  

Curiously, the basics are common across styles of cooking.  You have
to learn to coax the flavors out of the fresh ingredients and transform
them into the proper texture and finish.  Once you've mastered Italian
cooking, you may not be a top notch German cook, but you're probably
just a recipe or two away from being able to produce a very nice
German meal...

Databases have a certain similarity.  If heading an Oracle project and
I was given the choice between two people to work on my project, one
having been the lead architect for a top notch product based on
Sybase, and the other being an OCP that had worked on lack luster
products, it would be hard not to pick the former.



F> Following the same logic..... if I learn to cook a good Italian dish, then I
F> must automatically be an expert in preparing top-class Chinese, German,
F> Malay, Hungarian and French cuisine .... Yeah, right !

F> Ferenc Mantfeld