BI Accelerator Cube Reprocess Problems

We are currently trying to implement a BI solution for a data warehouse,
but can't seem to get the last few kinks worked out.
I have a cube made by BI that has monthly partitions. The problem is
that whenever I add new dimention information (which should be an
incremental update), I have to reprocess the entire cube. This is not
really an option, since it takes 57 hours to do that. So, does anyone
have a suggestion why BI is not doing an incremental update?
Thanks,
Chris
*** Sent via Developersdex http://www.codecomments.com ***
Don't just participate in USENET...get rewarded for it!
Two comments:
1) Re: "The problem is that whenever I add new dimention information (which
should be an incremental update), I have to reprocess the entire cube." Can
you be more specific what you are using? To "add" this new information are
you doing a full process of the dimension? Or an incremental process of the
dimension? I am a bit confused because you said "which should be an
incremental update" -- is it or not?
One of the things that you need to remember is that when you do a full
process of a non-changing dimension (regardless if the data changed or not),
then a full process of the cube is always required. This is because a full
process of the dimension could have caused changes in the hierarchy, which
would have likewise changed all of the aggregates -- which is a full
reprocess.
2) Re: "This is not really an option, since it takes 57 hours to do that."
You should be looking at improving that time. The general rule of thumb that
I use is that a server-class machine processing a typical partition should
be able to process about 1 million rows per minute. If your system is not
getting performance close to that then you should be looking at the general
best practices outlined in the SSAS Performance Guide at:
http://www.microsoft.com/technet/pro.../ansvcspg.mspx
see the section titled: "Optimizing Analysis Services to Improve Processing
Performance"
Dave Wickert [MSFT]
dwickert@.online.microsoft.com
Program Manager
BI SystemsTeam
SQL BI Product Unit (Analysis Services)
This posting is provided "AS IS" with no warranties, and confers no rights.
"Chris Timko" <ctimko@.hersheys.com> wrote in message
news:eP1Nk8lzEHA.2316@.TK2MSFTNGP15.phx.gbl...
> We are currently trying to implement a BI solution for a data warehouse,
> but can't seem to get the last few kinks worked out.
> I have a cube made by BI that has monthly partitions. The problem is
> that whenever I add new dimention information (which should be an
> incremental update), I have to reprocess the entire cube. This is not
> really an option, since it takes 57 hours to do that. So, does anyone
> have a suggestion why BI is not doing an incremental update?
> Thanks,
> Chris
> *** Sent via Developersdex http://www.codecomments.com ***
> Don't just participate in USENET...get rewarded for it!
|||It is simply adding new members to the lowest level of the hierarchy,
there were no changes being made to the rest of the structure. I guess
I am going to have to look at some of the other ways BI processes
dimentions.
And yes, we would like to get our cube running a little (actually, a
lot) faster. Right now, though, we need to make sure a full reprocess
isn't needed every time we add items.
Chris
*** Sent via Developersdex http://www.codecomments.com ***
Don't just participate in USENET...get rewarded for it!
|||Adding new members should *not* force a full reprocess.
Look in the documentation, I believe there is a way to force it do an
incremental only.
As a worst-case, you can modify the packages by-hand for only do an
incremental.
However I had thought that we had this worked out correctly so that we
detected if an incremental or full was required, e.g. deleting a member
forces a full.
Dave Wickert [MSFT]
dwickert@.online.microsoft.com
Program Manager
BI SystemsTeam
SQL BI Product Unit (Analysis Services)
This posting is provided "AS IS" with no warranties, and confers no rights.
"Chris Timko" <ctimko@.hersheys.com> wrote in message
news:eAe%233UL0EHA.3820@.TK2MSFTNGP11.phx.gbl...
> It is simply adding new members to the lowest level of the hierarchy,
> there were no changes being made to the rest of the structure. I guess
> I am going to have to look at some of the other ways BI processes
> dimentions.
> And yes, we would like to get our cube running a little (actually, a
> lot) faster. Right now, though, we need to make sure a full reprocess
> isn't needed every time we add items.
> Chris
> *** Sent via Developersdex http://www.codecomments.com ***
> Don't just participate in USENET...get rewarded for it!
|||I made the change to the only incremental update for the dimentions, and
it seemed to update just fine. If a full process needed to be run
instead of the incremental, would the BI dts update package crash? If
so, then it seems like there may exist a bug in the auto update of
dimentions. I plan to at some point (after some data massaging) use the
flagged update algorithm, so some of the dimentions can be changed.
thanks for the help,
Chris
*** Sent via Developersdex http://www.codecomments.com ***
Don't just participate in USENET...get rewarded for it!