Re: Convert SAP Oracle Database to IBM DB2 Database??

Noons wrote:
> Serge Rielau wrote:
>

>>> you know what is amazing in this paper?
>>> How it shows improvements in CPU in the application server
>>> when the db server gets compressed rows!
>>> Must be an amzing "feature", this compression that
>>> can change the performance in a different system...

>> What are you referring to? Care to share a page number?

> What I'm talking about is that paper making
> false claims about db CPU use reduction by measuring
> application dialogue CPU use.

It does? Where does it say that it is measuring application dialog CPU use? It's measuring three values:
CPU time
  Which consistently is SERVER.
  While that is not spelled out explicitly all of section 6.3 is pretty   clear about it.
Response Time
  That's bumper to bumper including the app DB Time

   Wall clock time consumed on the server    (I hope you take no issue with CPU time > Wall clock time)

>
> SAP is a multi-tier application-server based system.
> If you change the CPU usage in the database it works against,
> you change NOTHING in the CPU usage of the application
> server in which SAP is being executed. Note that I am talking
> about CPU Usage, the term used in that paper. NOT overall
> wall clock time or duration.

I agree. But what substantiates your claim that this is what it says? It is unfortunate that the paper assumed a reader who doesn't actually try to misunderstand the obvious definition for the purpose of discrediting it.

> The paper contains multiple examples near the end
> of " dialogue test cases" in which the " CPU usage" by
> SAP was reduced.

By CPU Usage SAP? Where does it say that? CPU Usage has not been redefined. It merely correlates the data previously collected on the server to specific dialogs.
Please, perhaps my English is really bad.. point to your source so I can learn. Don't leave me ignorant.

> Yeah, I'm sure this was caused by DB2's compression.
> And the pope is not a Catholic.

Teh server side CPU changes, yes. No claims made on the app.

> Talk about NO CLUE, from whoever lousy marketeer
> pushed this one out...

I took the liberty of checking out the author. Just as the keyword "PRODUCTION" in the title might suggest he is indeed a DBA.
This is "his" system. Note that he is looking at long term results, not some 3 hours run. The graph itself covers a month.

Cheers
Serge

-- 
Serge Rielau
DB2 Solutions Development
IBM Toronto Lab