Using TM1 over Wide Are Networks

[mce]

Hi,

As I am new in TM1, I have a question for the experts:
- I heard that TM1 is not performing well in general over wide area networks due to its dependence on server processing.
- In case of multinational companies having TM1 users accross different countries, what is the best practice to use TM1? Is it through TM1 Web, TM1 Contributor, or through Citrix...etc?
- Which client tool is performing better in such distant use of TM1 amongh TM1 Perspectives, TM1 Web, and TM1 Contributor?

Thanks in advance for replies.

Regards,

[George Regateiro]

You can search to see if there are more posts here are just a few

http://forums.olapforums.com/viewtopic. ... rix#p13490

http://forums.olapforums.com/viewtopic. ... lit=Citrix

The main thing to think about when evaluating a solution is what functionality do you need to expose to remote users. Do they need the full TM1 worksheets or can they handle smaller entry and reporting Templates? Are you using VBA? Can you take a multi tiered approach (Citrix and Web). Unfortunately there are no hard and fast answers to the question, it is really usage/budget dependent.

[Martin Ryan]

The reason TM1 is slow over the WAN is not because of it's reliance on server processing, but because it has a very chatty protocol between the server and the client, which means a fair amount of bandwidth between the two is required.

TM1 Contributor looks to be where IBM are putting their time in terms of addressing this weakness of TM1. Which is understandable but a bit frustrating for those who have always enjoyed the Excel interface.

Contributor is web delivered, so no installs on clients' machines. It should also be much faster than those chatty protocols that are required when using the cube viewer or Excel over a WAN.

TM1 Web is also the other option and allows a bit more flexibility to the user, e.g. creating their own views.

Martin

[mce]

Thanks for the replies.

The reason TM1 is slow over the WAN is not because of it's reliance on server processing, but because it has a very chatty protocol between the server and the client, which means a fair amount of bandwidth between the two is required.

I was looking from a higher perspective that the client might load the data and the model when first opening, and then it may not need to communicate to the server until the user decides to save the changes to the server, in which case the client would not be relient on server to process the calculations and display the results as the calculation could take place on the client side as well since both the model and the data is available in the client machine. This is the case on Contributor of Cognos Planning, but not in TM1 web & contributor clients and therefore there is a need in TM1 to continuesly communicate to the server as the user makes entries or change the view layout.

On the other hand, it looks you are right that if the network protocol of TM1 was not chatty, it might have been performming better over WAN.

Regards,

[Martin Ryan]

then it may not need to communicate to the server until the user decides to save the changes to the server

That's not my understanding of how I read the doco and would represent a major change in TM1 architecture because all the rules would have to be downloaded to the client. The way TM1 works I think that would be impossible.

I have also noticed that there are sandbox logs on the server which would indicate the data is flowing back to the server before commitment.

Martin

[David Usherwood]

Planning pulls all data back to the client and models it there. TM1 sends all changes to the server where the calculations are performed and the results delivered back. Both approaches have pluses and minuses.
From the Black Belt sessions in July I gather that Iboglix want to move towards bringing some calculation logic to the client if required and ultimately to support disconnected data and calculations. Don't think it's a half hour fix though (to put it very very mildly indeed).

[mce]

That's not my understanding of how I read the doco and would represent a major change in TM1 architecture because all the rules would have to be downloaded to the client. The way TM1 works I think that would be impossible.

I did not mean TM1 should or will have such feature, but only tried to idendify the architectural feature of TM1 that requires continous communication between the client and the server.

However I also heard that they are trying to add work-offline feature to TM1 Contributor (as we have it in Cognos Planning Contributor), but is not planned to be available in the upcoming release.

[mattgoff]

Planning pulls all data back to the client and models it there. TM1 sends all changes to the server where the calculations are performed and the results delivered back. Both approaches have pluses and minuses.
From the Black Belt sessions in July I gather that Iboglix want to move towards bringing some calculation logic to the client if required and ultimately to support disconnected data and calculations. Don't think it's a half hour fix though (to put it very very mildly indeed).

Interesting change, but I wish they would work on some more basic fixes first, like filling up the entire TCP payload instead of sending one TM1 "unit" (element name, cell value, etc) per packet. Bonus points for not requiring an ACK before sending the next packet-- just burst them all out asynchronously and resend the missing ones at the end....