Hello

I have a really strange problem. I am using the Business Intelligence Portal
2003 with a customer to browse their OLAP cube. The performance of the
application has been quite good until the other day, when all of a sudden it
is incredibly slow.

Some network analysis showed me that tha OWC components (two pivots and on
chart in each view) seem to download about 6 mb each from the server
everytime they are executed. I do not know if they did that before, I only
know the application used to be fast and now it isn't.

I thought OWC cached the data? Why does it need to download all the data
everytime? Does anyone know?

Is it possible to force it to cache?

Thanks!

Best regards, Ola

do you have any calculated measure?
do you use AS2000 or AS2005?
what is you connectionstring?

for AS2000, verify that the client computer has the drivers from the latest
SP. (install the latest PTSLITE.exe)

does the problem also occurs when the user simply access the cube from any
other method? (like from Excel)


"Ola Larsson, Fishbone" <OlaLarssonFishbone (AT) discussions (DOT) microsoft.com> wrote
in message news:A80B7C7E-BA57-4A66-94EB-5037E520F7CD (AT) microsoft (DOT) com...

Hello Jéjé

Yes I have six calculated measures, but I am only using three of them at a
time.
We are using AS2000. I do not know what version of OWC the clients are
using, however they all have Office 2003...

Generally it seems that if clients are accessing the cube directly from
excel, it is lightning fast.

Thanks!

Ola


"Jéjé" wrote:

having OWC is not enough, you have to confirm the version of the OLAP
driver.
to insure that the client has the right version, install the PTSlite.exe
application (from the SP4) on a desktop to install the latest OLAP drivers.

what are your formula in the calulated measures?
have you setup the non empty behavior option?

have you identified if the problem come from 1 measure only? (remove all
measures except 1 and see the result)

if you want to execute the query on the server side, add this in the
connection string:
Execution Location=3
but first, try to add "Cache Policy=7" in the conneciton string (SP4 driver
version required)

But make sure you test the results and after correctly!!!
In our case we suffer issues when the formula is executed on the server side
so everything run on the client side.

"Ola Larsson, Fishbone" <OlaLarssonFishbone (AT) discussions (DOT) microsoft.com> wrote
in message news:CB177835-CB20-41BF-8F48-666D11EB3057 (AT) microsoft (DOT) com...

Hello Jéjé

Ok, I installed the lates version of PTSLite an aclient. I removed all
calculated measures from the cube, set Cache Policy = 7 and tried the
Execution Location = 3, but it seemed to not make a difference.

The client still downloads about 18 mb (3 OWC pivots) in each view.
Strangely enough the system was really fast just a week ago, and I cant think
of anything that has changed...

Any other thoughts?

Best regards, Ola

"Jéjé" wrote:

Hello Jéje´

Ok, I installed the PTSLite on a client. Cganged the connstring, adding
Cahce Policy = 7 and tried the Execution Location = 3. I even removed all
calculated measured from the cube.

The client still downloads about 6mb for each OWC in a view, regardless. The
strange thing is that just a week ago, it behaved differently...

Any other thoughts?

Best regards, Ola

"Jéjé" wrote:

have you executed the usage based optimization wizard? (maybe the system
retrieve lower levels information due to a lack of aggregated values)

and do you do a full process of the cubes & dimensions?

during the week, have you noticed a changes in the size of your dimensions?
(like a huge new list of products)
with AS2000, a lot of metadata must be sent to the end user, which result in
a lot of traffic only to transfert this information.
Also, if you displaay a dimension like "customers" and you display only the
top 10, AS2000 send more customers data other the network because the
formula is executed on the client side.

Try other things in the connectionstring:
Cache Ratio=0.01;Cache Ratio2=0.01;Cache Policy=7;Non Empty
Threshold=1;Large Level Threshold=100

The cache ratio2 option inform AS on when and how many NON EMPTY members are
retrieve from the server. If AS detect more then XXX members with data, then
all the members are returned, if its less then the members are returned 1 by
1.
Large Level Threshold also help the system to analyze where the query should
run (if more then 100 members are requested, then the query is executed on
the server side)

or this:
Execution Location=3;Default Isolation Mode=1


"Ola Larsson, Fishbone" <OlaLarssonFishbone (AT) discussions (DOT) microsoft.com> wrote
in message news:5B2CFE5A-D572-4353-86E0-01F437434F14 (AT) microsoft (DOT) com...

Hello Jéjé

It doesn't seem like anything I do improves the performance or decrease the
amount of data downloaded to the client.

For example if I define three empty pivot-OWCs using the same connstring as
the others they still download 6mb each.

I do not quite understand what it is they are downloading...

best regards, Ola

"Jéjé" wrote:

the system send the dimension list, levels list, etc...
more dimensions you have, more data has to be sent other the net.


"Ola Larsson, Fishbone" <OlaLarssonFishbone (AT) discussions (DOT) microsoft.com> wrote
in message news:A3415CB2-B6C3-424C-B877-C7CA373ED8D2 (AT) microsoft (DOT) com...

Hello Jéjé

Thank you for replying, still ;-)

I made a performance comparison to the Foodmart Warehouse and Sales cube,
set up two OWC pivots and one chart in a view, and it is about the same thing
there. I do not think this is really an OLAP issue. It seems more like a
connectionstring or IIS issue.

Best regards, Ola

"Jéjé" wrote: