OmniBase Performance..

Here is an interesting test.

repo := 'C:\Benchmark'.

(OmniBase createOn: repo) close.

time:= Time millisecondsToRun: [
Transcript cr.
db := OmniBase openOn: repo.
[OmniBase root at: 'Sequences' put: (Dictionary new)]
      evaluateAndCommitIn: db newTransaction.
[base := OmniBase root at: 'Sequences'.
base at: 'CustomerNo' put: 0.
base markDirty.
] evaluateAndCommitIn: db newTransaction.
[0 to: 999 do:
      [:i|
      (i \\ 100) = 0 ifTrue: [Transcript show: '.'].
      [base := OmniBase root at: 'Sequences'.
      nextCustomerNo := (base at: 'CustomerNo') + 1.
      base at: 'CustomerNo' put: nextCustomerNo.
      base markDirty] evaluateAndCommitIn: db newTransaction.
      ]] ensure: [db close]].

Transcript cr; show: (time/1000) asFloat displayString; show: ' seconds'.

db := OmniBase openOn: repo.
base := (db newTransaction) root at: 'Sequences'.
Transcript cr; show: 'Last Customer No: '; show: (base at:
'CustomerNo') displayString.
db close.


Now, the performance is this:

On my desktop machine, AMD XP 2500/1GB, This is executed in about 2.5
seconds (yay)

On my work machine, P4 2.8/1GB, this is executed in about 75 seconds. (not
yay)

On my laptop, P4 1.6/512MB, it is executed in about 110 seconds.


There is no other difference really. They're all running OmniBase 2.1, all
running in a fresh 5.1.4 image, all on XP SP2. The only obvious difference
between them is the CPU (And drivers, motherboards, oh, well pretty much
everything I guess)..

So I'm think that perhaps maybe it's the motherbard chipset/disk io
controller? the AMD box has an nforce2 chipset, and the P4 machines have
standard intel p4 chipsets (whatever the name of that is.. )

Anyone experienced such dramatic differences in disk IO when changing
between CPUs? Is there some driver update I might be required to get decent
performance out of Intel machines?

My non-scientific experience is that my AMD machine blows the Intel machines
out of the water, and not just in Dolphin.

Also,

Take this code for example.

Time millisecondsToRun: [
 1 to: 1000 do: [ :i | (OmniBase openOn: ('C:\Sys\Database_%1!i!'
formatWith: i)) close ] ]

(The DB's already exist)

On the AMD machine, run time: about 3 seconds.

Intel machine, over a minute.


I' really thinking it's a driver issue. The AMD machine is using an nForce
based motherboard and the nForce based chipset drivers. The Intel machines
are using stock standard Windows XP drivers.

I installed the Intel chipset driver into the laptop and shaved the
execution time by 10 seconds (100 seconds down from 110 seconds)

I'd be interested to know what other peoples performance results are, and
the machines/driver combinations involved

Sean M,

I run it:

Time: 65 seconds on:
P4 2.4 Win XP SP2.
512 RAM.

Interesting results.
Do you know anything about this nforce2 chipset ?
May be in the manuals there's an explanation about disk IO.

Regards Bruno

If all the test where made in localhost (as C:\benchmark) I think the
hard disk performance should be a critical parameter for those tests.
So same CPU-mother-chipset could give very different results with
diferent hard disks.

regards,

Sebastián