Re: Can someone suggest me how do you define a big cube..

Tech Tip: Click here to run a free scan for Windows Errors and optimize PC performance

oh... AS2000...
yes, no profiler for AS2000.
and no attributes in AS2000.

if you have a DCount aggregated measure, you must store it in a dedicated cube to make sure the performance was good enough.

to improve the performance insure that your hierarchies in the dimensions are setup with the uniquekeys option. well, each level must use a unique key in this case, so a good dimension design is required, but the impact on the performance can be important.

does your slowest measure is a calculated measure?
if yes, send the code please.


"demonhead" <ravi0219@xxxxxxxxx> wrote in message news:1175588242.714911.322560@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
On Apr 3, 4:47 am, "Jeje" <willg...@xxxxxxxxxxx> wrote:
How the memory is used on the server?
if your overall metadata is large (a lot of attributes) this takes some
memory.

Does your queries hit your aggregated content? (use the profiler)
what is your front end tool?

do you have distinct count measures? if yes, do you store this DCount in a
dedicated measure group.
how many rows do you have in your fact table? do you use partitions?

does your attributes use the rigid relationship option?

what is your disk subsystem? raid 5, 1+0? SCSI, SATA, SAS?
and how big are your aggregations and MOLAP storage on these disks?

good luck!

"demonhead" <ravi0...@xxxxxxxxx> wrote in message

news:1175531050.658489.228820@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx



>I will depict my current scenario and someone can comment whether my
> cube is big or not...
> I have 24 dimensions and 2 measures.
> My fact table can be said to have over 3 lakh records...
> About 3 dimensions have 20000 members plus and 2 dimensions have more
> that 5k members..
> I cannot remodel those 5 dimensions to have hierarchies.
> When i have a top-in-top concerning the any of the three dimension in
> the inner top and about three members in my where the query just stops
> performing all the performance strategies of partitioning i have tried
> out but they dont work...
> Can anyone suggest me performance boosters for my current scenario,,
> anything leading from design storage (Currently the cube is molap and
> have designed aggregations based on the usage optimization)..
> Anybody can please share their experiences ... for cubes same as these
> sizes or even larger ones...
> I want the response times to be around 30-35 secs for the heavier
> queries.. and not much more

> I am currently using AS2K(32-bit) on win2003 box with Xeon 3.2
> processor, and 4GB Ram..

> Help would be appreciated

> Thanks and regards- Hide quoted text -

- Show quoted text -

hi Jeje,
To start with disk subsystem is Pata drives...
I do have distinct count measures but i do not store them in dedicated
measures group
I do not use attributes at all..
Aggregation storage space currently has been kept upto 100MB
I remade all the aggregations after firing the queries once but do not
feel that queries are resolved from aggregations because they take the
same amount of time..
In AS2K i believe that profiler can only be used for ROLAP cubes and
not MOLAP.
I have 3 lakh rows in facttable, and about partitions i wanted to ask
one thing would partitions be effective only on dimension members
that i have used in where clause of my MDX queries

.

Relevant Pages

  • Re: Can someone suggest me how do you define a big cube..
    ... Does your queries hit your aggregated content? ... and how big are your aggregations and MOLAP storage on these disks? ... I have 24 dimensions and 2 measures. ... About 3 dimensions have 20000 members plus and 2 dimensions have more ...
    (microsoft.public.sqlserver.olap)
  • Re: SSAS2005 - When do partitions need to be processed?
    ... My guess here is that it is not aggregations but rather indexes being ... The actual processing of partitions doesn't appear to be significant here ... Processing Cube 'Sales Current' completed successfully. ... aggregations when the dimension is updated. ...
    (microsoft.public.sqlserver.olap)
  • Re: Aggregations/performance
    ... If you get good performance out of your cube, ... The first point is that you can certainly have aggregations with a changing ... As a worst-case, even if you have several changing dimensions, this ...
    (microsoft.public.sqlserver.olap)
  • Re: dimensions and cube aggregs
    ... You might consider looking at the Analysis Services Performance Guide. ... > I have 8 dimensions that are needed for various queries. ... > I am trying to use them all in 1 cube, ... > The problem is the aggregations get a bit out of hand even at 20%. ...
    (microsoft.public.sqlserver.olap)
  • Re: Can someone suggest me how do you define a big cube..
    ... Does your queries hit your aggregated content? ... and how big are your aggregations and MOLAP storage on these disks? ... > cube is big or not... ... > I cannot remodel those 5 dimensions to have hierarchies. ...
    (microsoft.public.sqlserver.olap)