Cognos Planning 10.1 Server Specification

[Hector]

Hi,

We currently use EP 10.1 but Analyst only. We want to introduce Contributor on a small scale (5-10 Users).

I have searched for the hardware specification for a server. I assume I need a different server to the one that runs Analyst and it needs to be a web server. Does anyone know where I can find the recommended server specification?

Many thanks,

H

[ericlfg]

Hi Hector,

Depending on the current hardware and based on the number of users you have planned, you might be fine with the server you have.  It is recommended to not use a single server where applicable, but certainly it's not always required.

What are the specs of the current server hardware?

[Hector]

Hi Ericlfg,

Thanks. Below is our current specification.

CORES: 2
RAM: 4GB
Processor: Zeon 2.13 GHZ

[ericlfg]

Hey Hector,

So thinking a little more about this, the question of the server specifications is a bit of a tricky topic.  There are some performance aspects as well as modeling aspects that you may never have focused on due to being an Analyst only shop.  The 2 primary aspects would be:

1. Size of the Analyst models (largest cube, largest dimension, etc.) as well as what kind of modifications may be required so it can be used in Contributor.
2. Number of elist items (a term you will want to get familiar with -- at a high level this is how your organization will capture contributor data.  Ex: Geographic cost centers, projects, manufacturing aspects, etc).

When building an environment for use with Contributor, you need to be aware of the size of the model and the number of elist items as each elist item will often go through a Reconcile process.  The reconcile is triggered by updates to the Analyst model that are then synchronized with the Contributor applications.  Having a single server with 2 cores may yield poor Reconcile performance (the time it takes for each node to reconcile multiplied by the total number of elist items), as well as poor performance for end users while the Reconcile is running (the job servers are designed to utilize 100% of the available CPU (2 cores = 50% utilization each)).  On a single server all aspects will compete for processor time and thus reduce the total performance.  If you will not be reconciling often, then this is a fairly moot topic; however, if you make regular changes to an analyst model and this model will be used in Contributor, these topics are important.

Long story short, you may want to engage with a consulting firm to evaluate your current analyst models to determine if they're appropriate 'out-of-the-box' for Contributor, as well as evaluate your server environment to see if, in it's current state, it will yield acceptable results.  At a minimum I would probably recommend a separate job server machine (the specs would depend on the 2 variables mentioned above).  Having a separate job machine allows processing to be performed without affecting any of the core responsibilities the current server would perform (Content Manager, Gateway, Planning web service, CAC service).

Cheers

[Hector]

That is a great response. I think our current Models will not suit contributor at the moment, which is why I plan to try it on a small scale on some stand alone Models.

I will get in touch with a consultancy and see how it goes.

Thank you

[kakkaduan]

Hi,

Excellent post by Ericlfg, one more thing:

What triggers a reconcile is also updates to the data within the model.

I'm saying this so that you don't base your decision on number of servers on something like "Our model(s) is/are static so reconciles won't be a problem". I've seen environments where plain updates of data is causing the reconcile to processes fail or take almost forever. So basically a static model with big/frequent data updates might also be an incentive to add additional job servers to the environment.

PS. Ericlfg: Sorry if by "model" you were referring to both the model structure and the data within it, In my world however these are two separate things.

BR
Jakob

[ericlfg]

Hey Jakob,

You are absolutely correct and I wasn't really considering that side of things; I should have expanded on the reconcile topic, as it can be invoked by a number of different processes.  EX: Admin links, C<A d-links, Import data, Access Table changes.. 

I also consider the data within the model and the model itself as 2 separate entities.  They really do have to be considered separately since you can have a model without data, but you can't have data without a model..  Not sure how either would be beneficial, but I think you know what I mean..   :)