Trevor

If you are bored, the real test is a mutidimensional file ie Invoice and
Items and compare that to a select against a joined table on SQL Server
particularly where you select criteria from both tables.  Then for fun
add record locking with say a 1000 routines locking and updating records
at a time.   And watch what happens to the performance when SQL Server
trys row locking and/or pessamistic locking.  Of course you could
probably set that up in Universe over  night, but will need a couple of
days to set up the other side.

Joe has to be a bit fair on his analysis, the Universe System is
probably handling all the data entry, batch process, record locking and
updating with a large number of users, where the SQL Server is acting as
a data-warehouse with less users and no updating.  Does the CIO of his
company agree with his analysis.

I have run similar application on Universe and SQL Server that involved
multiple selects and lots of updates with transaction boundries and
found Universe to be a fractional faster.  (There was no tuning done to
either database)

Just some areas that I have experience problems with Oracle that would
have been less in Universe.  (design also plays a role)
On the RTA system with Oracle, they had normalised to the point of
requiring 19 joins to select certain criteria.  We found definite
degradation as we moved over 3 or 4 joins.  Multidimensional design
would have minimised the joins.

There was at that stage a requirement to restrict the size of Oracle
database.  So a large Database had to run over multiple Database
Partitions.  When you exceeded two partitions performance dropped.  Of
course improvements in Oracle have reduced this issue.

One bureau service would not let me load ODBC on Universe, because their
bad experience with an RDBMS (wont mention for political reasons) where
ODBC Queries brought the system to its knees.

The biggest issues with RDBMS I have seen over the years, is that all
the fancy tools do great things with enquiring the database and they are
"WOW", but updating is very complex and more times than often have seen
programmers stuff this up with performance degradation and even loss of
integrity.  Its all easy to write an update command in VB, but it's the
setup of the cursor than can be critical.  (They don't call it
optomistic locking for nothing, because that is all it is pure
optimism).


Joe, we are not amateurs, many of us have been involved in multinational
companies, processing billions of dollars of transactions in mission
critical from banking to health to distribution.

Regards

David Jordan

"God gave us two ears and one mouth so that we would listen twice as
much as we talk"








-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Trevor Ockenden
Sent: Wednesday, 31 March 2004 4:19 PM
To: U2 Users Discussion List
Subject: Re: Comparing Oracle with UV


Joe

I do not have access to a file with 20 million records but I will create
one with random data and try to carry out your test. Bare in mind I will
be using my laptop (NEC Versa LX P2 366 - 384Mb RAM and relatively slow
disk) which may sku the results somewhat but I will do my best to tune
UV to get satisfactory results.

With luck and in between picking up the kids etc. I should have a result
by tomorrow (as it is nearly knock off time Down Under).

I would be interested to know if you could create the environment from
scratch to carry out this test as I am intending to do? How long would
that take you?

For your interest, whilst I was typing this e-mail my program to create
the data has already created 1 million records. It took approximately 5
minutes to write. Compiled first time. I also have [EMAIL PROTECTED] running,
plus 10 other windows open. Total memory usage is 244Mb on Win2000.

Cheers

Trevor Ockenden
Open Systems Professionals

----- Original Message ----- 
From: "Joe Eugene" <[EMAIL PROTECTED]>
To: "U2 Users Discussion List" <[EMAIL PROTECTED]>
Sent: Wednesday, March 31, 2004 3:30 PM
Subject: RE: Comparing Oracle with UV


>
> Trevor,
>
> I will try to do this "ANALYZE.FILE STATS" sometime tomorrow. Again...

> Nobody here has come up with any real proof of any clear testing 
> results...other than just sending out useless random emails.
>
> I provided the exact details on my testing, can you do me a BIG Favor.

> Do you have any UV DataBase that Contains around 20 Million Records?
>
> Can you do a CASE IN-SENSITIVE Search against ONE Field in UV 
> FILE/TABLE and post the real time average results?
>
> I belive the syntax UV uses for WILD CARD Search is "]"
>
> I simply cannot even come with a Seconds Timing againt less than 1 
> Million Records in MS-SQL Server with NO Indexes, which proves very 
> good performance.
>
> I have never seen this Kinda Good Performance in any UV 
> Programs/Against
any
> UV Files
> within UV/PICK/BASIC.
>
> Thanks,
> Joe Eugene
>
>
> >-----Original Message-----
> >From: [EMAIL PROTECTED] 
> >[mailto:[EMAIL PROTECTED]
> >Behalf Of Trevor Ockenden
> >Sent: Tuesday, March 30, 2004 11:26 PM
> >To: [EMAIL PROTECTED]
> >Subject: Comparing Oracle with UV
> >
> >
> >Joe
> >
> >As this comparison is still raging why don't you do the ANALYZE.FILE 
> >STATS for us and post the results as my experience has shown that 9 
> >times out of 10 performance is related to file sizing. A fundamental 
> >element of setting up any database.
> >
> >Note, we all stand to learn something here.
> >
> >Cheers
> >
> >Trevor Ockenden
> >Open Systems Professionals
> >M: +61 414 731 634
> >E: [EMAIL PROTECTED]
> >
> >
> >---
> >Outgoing mail is certified Virus Free by AVG 6.0.
> >Checked by AVG anti-virus system (http://www.grisoft.com).
> >Version: 6.0.642 / Virus Database: 410 - Release Date: 25/03/2004
> >--
> >u2-users mailing list
> >[EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users
>
> --
> u2-users mailing list
> [EMAIL PROTECTED]
> http://www.oliver.com/mailman/listinfo/u2-users
>


---
Outgoing mail is certified Virus Free by AVG 6.0.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.642 / Virus Database: 410 - Release Date: 25/03/2004

-- 
u2-users mailing list
[EMAIL PROTECTED]
http://www.oliver.com/mailman/listinfo/u2-users

-- 
u2-users mailing list
[EMAIL PROTECTED]
http://www.oliver.com/mailman/listinfo/u2-users

Reply via email to