In AS2000 DSO it was possible to create a 'temporary' partition whose name
began with a tilda e.g. ~Partition1, which then did not participate in the
cube until it was merged with a normal partition. Can this be done in AS2005
? I've tried creating a partition beginnign with a tilda through AMO but it
just seems to appear as a normal partition. I'd like to build some partitions
'offline' and merge them in after all are complete. Any ideas of how to
achieve this or something equivalent would be appreciated.

I'm not sure if I understand your needs...
but AS2005, like AS2000, use transactions, so the partition become online
when the process is finished.
so what are you looking for?

"Steve G" <SteveG (AT) discussions (DOT) microsoft.com> wrote

I've got several partitions and they take a long time to build. I'd like to
process them all, then switch them 'on'. I don't want to use an AMO
transaction because all the processing will be rolled back if one partition
fails (which is quite possible) - and I'd have to start again from the
beginning - wasting a lot of time. If I get a failure I want to fix the
problem in the source data and then carry on processing partitions from where
I left off, then switch on all the partitions at the end (otherwîse users see
some partitions updated and some not - which is bad). This seems like a
reasonable operation to me and it was possible in AS2000 because it was
possible to create 'temporary partitions'. Now I don't see how to do it in
AS2005, do you ?


"Jeje" wrote:

ok,

try this:
install 2 instances of AS2005
do the process on the instance A, when is finish, synchronize instance B
with instance A.
it could be 2 different servers or 2 instances on the same server.

"Steve G" <SteveG (AT) discussions (DOT) microsoft.com> wrote

Using a seperate server instance is not practical because all my users are
attached to instance X from where they run reports. If I synch instance Y
they'd all have to change their connections every time I synch (daily), which
isn't practical. I need the copy to be on the SAME SERVER. Also, multiple
instances means a more complex environment to maintain (unnecessarily so in
my opinion).

To me it looks like a short-sighted omission by MS. I've decided to use the
backup/restore functionality although it is very slow (and I guess I'll lose
all cached data daily since the db will be completely replaced each day).


"Jeje" wrote:

no, when I say "Synchronize" this step "simply" copy the cubes from instance
A to instance B.
Look at the "Synchronize database" option in the management studio. This
command can be executed using XMLA commands, so you can easely schedule this
step in your daily process.
the users continue to use the same connection, there is no change for them.

its not a complex environment its an environment where you want to process
your cubes without any impact on the users.
in our case we have dev, test and prod environments and switching from 1 to
another is really easy, quick and simple. there is no administration
overhead.

your backup / restore option is like the synchronize command, but its not a
beautyfull solution compared to the synchronization.

You talk about loosing the cache. you loose this cache when you process a
cube or you do a synchronization.
To fill the cache again, I recommand to execute some MDX queries against the
new cube to fill the cache.

"Steve G" <SteveG (AT) discussions (DOT) microsoft.com> wrote