TM1 Forum

I have three servers in my "sandbox". I am in the process of converting from Vers 6 to Vers 9 and am learning Turbo Integrator.

I need to write a TI process where the data source is a cube view on A: and one-off data is collected from cubes on B: and C:

Essentially, I want something that looks like this:

data source on ServerA, process on Server A

VariableB = cellgets('b:cubeB, elements etc);
VariableC = cellgets('c:cubeC', elements etc);

Can this be done?

I have tried expressing the cubes in UNC form:

\\sandbox\tm1\a\data\cubename

but that does not seem to work.

Help. Please. I tried searching the forum first, but could not find anything.

Carolyn wrote:I have three servers in my "sandbox". I am in the process of converting from Vers 6 to Vers 9 and am learning Turbo Integrator.

I need to write a TI process where the data source is a cube view on A: and one-off data is collected from cubes on B: and C:

Essentially, I want something that looks like this:

data source on ServerA, process on Server A

VariableB = cellgets('b:cubeB, elements etc);
VariableC = cellgets('c:cubeC', elements etc);

Can this be done?

I have tried expressing the cubes in UNC form:

\\sandbox\tm1\a\data\cubename

but that does not seem to work.

Help. Please. I tried searching the forum first, but could not find anything.

Sorry, I'm afraid that server-to-server communication is not readily done in TI. (But should be... that, however, is another gripe (which IBM will ignore) for another time.) There are a few ways around it. I was going to write them out but found that David Usherwood had already done so in this thread.

One option that he didn't mention if this is a one-off is to copy the cubes on the other server across to the main one (either by physically copying the cube, rule and dimension files) or by replication. After you're done you can just delete the one-off cubes.

(But should be... that, however, is another gripe (which IBM will ignore) for another time.)

Yeah, it should be, but is not. Sigh.

Since I'm still in the sandbox, it looks like I'll have to move it all to one server. Snarl. Gripe.

Thank, again, Alan. I really appreciate your help.

You can move it all into one server, but many TM1 shops work with a farm of servers and move data around with flat files - see my post referred to by Alan. Until 9.5.2 ParallelInteraction settles down and starts working as designed it can be a very effective way of allowing users continuing access to data while heavy calculations are running, especially with modern servers with lots of cores. But if you feel like raising an enhancement request then please do - it should add to the pressure on IBM.

You should be able to move data between same version TM1 servers by using the ODBO Connection datasource option within TI if your on Windows Servers. Select the ODBO Option and use the following parameters:
ODBO Provider: IBM Cognos TM1 OLE MD Provider
ODBO Location: b's Adminhost
ODBO Datasource: b
ODBO Catalog: b
ODBO UserID: whatever
ODBO Password: whatever

Then use an MDX Query or ODBO Cube, I just ran a test with a MDX Query between two servers on 9.5.2 seems to work fine.
This should work for you unless you will be connecting to version 6 servers or something crazy like that.

Oh and it used to be that if you used ODBO as a provider and connected to the server which is holding the TI your writing it would crash the server, but that seems to be fixed in 9.5.2 - another win for 9.5.

Kyro: same version TM1 servers by using the ODBO Connection datasource option within TI

That could work, except that I am already using a cube view as the data source.

The data source has a project code. I want to look at another cube on another server to see what type of project that particular project code has.

The company is involved in homebuilding. The data source cube view refers to "ABC" project, and has various data for ABC.

I need to look into another "project info" cube, on another server, to see if ABC is in a certain geographical area, is open or closed, grab the date of project inception, etc. Those initializing data points are used in the calculation that comes next.

I am now inclined to abandon the multi-server model and go to a single server. The three-server model was set up in our old Version 6 and I had just duplicated it in the Version 9 sandbox. I believe it was set up that way for ease of use in Version 6, since the Version 6 cube pick window is so tiny. I am not convinced that the need for three servers overweighs the hassle I am seeing now. Since I'm in the development / sandbox stage, I can make a radical change without freaking out the user community.

Question: If I decide to move everything to one server, can I just copy and paste the relative files, then restart? Is it as simple as that? I have cubes, views, rules, TI processes and chores.

Carolyn wrote:Question: If I decide to move everything to one server, can I just copy and paste the relative files, then restart? Is it as simple as that? I have cubes, views, rules, TI processes and chores.

It should work fine, provided that there are no (different) objects with the same name that would overwrite each other. And, of course, provided that you copy all the object files that other objects depend on.

It should work fine, provided that there are no (different) objects with the same name that would overwrite each other. And, of course, provided that you copy all the object files that other objects depend on.

OK, good. There are no duplicate object names and I'll make sure I copy everything.

This seems to be the simplest solution to the problem, now and for similarl issues in the future.