placement of Analysis service? on cluster or lone server?

Hi, we have an active/passive cluster which hosts our production database.
Transactional replication is being made from the production to a 'reporting'
database residing on a lone server. We will need to install Analysis service
for the construction of cubes.
I was wondering where Analysis service should be installed?
Is is ok to install it on the lone server and have it use the data from the
reporting database (which is filled via transactional replication)? Will
Analysis service have a problem with this based on the fact that it is
reading a read only database which doesnt have primary keys in the tables?
OR, should analysis service be installed on the cluster? Is it cluster
aware?
At this stage we are thinking of only running the cubes at night to generate
the data for export to reporting service.
Would it then be a good idea to install reporting service on this same lone
server?
Any insight would be most appreciated!
Cheers, john
If that is the case, you can't possibly be running transactional
replication.
The "reporting" database is LOGICALLY read-only, not physically read-only.
This database also MUST have all of the primary keys in place on the tables,
since it is impossible to replicate a table with transactional replication
if the table does not have a primary key. This primary key exists at both
the publisher and the subscriber.
Analysis Services is not going to write to your tables when building cubes,
so it does not care if the tables are physically read-only, logically
read-only, or read-write.
Where you place AS is entirely up to you. You can install it on the
stand-alone server and you can install it in the cluster. The question you
need to ask is what level of availability do you need for AS. If it needs
to be available as much as possible, even through a hardware failure, then
it should be installed in the cluster. If not, it should be installed on
the stand-alone machine.
Mike
Mentor
Solid Quality Learning
http://www.solidqualitylearning.com
"john clarke" <jclarke@.nospam.com> wrote in message
news:%23QyI71JDGHA.688@.TK2MSFTNGP11.phx.gbl...
> Hi, we have an active/passive cluster which hosts our production database.
> Transactional replication is being made from the production to a
> 'reporting'
> database residing on a lone server. We will need to install Analysis
> service
> for the construction of cubes.
> I was wondering where Analysis service should be installed?
> Is is ok to install it on the lone server and have it use the data from
> the
> reporting database (which is filled via transactional replication)? Will
> Analysis service have a problem with this based on the fact that it is
> reading a read only database which doesnt have primary keys in the tables?
> OR, should analysis service be installed on the cluster? Is it cluster
> aware?
> At this stage we are thinking of only running the cubes at night to
> generate
> the data for export to reporting service.
> Would it then be a good idea to install reporting service on this same
> lone
> server?
> Any insight would be most appreciated!
> Cheers, john
>
>