Topology question for PowerPivot for SharePoint engine

Question

We currently have a SP2013 farm with two web front end servers, two application servers and a SQL 2012SP1 SQL Cluster.  We plan to expand the farm with all BI functionalities by adding another application server to SP2013 farm, which will be dedicated
to all BI services, such as Excel service, PowerPivot, SSRS integrated mode, PerformancePoint, etc. and two Database servers (belong to DBA team): one of which will be used for Data Warehouse, another one for SSAS instances, including Multidemensional Cube
instance, Tabular mode instance. My question is where is the best place for PowerPivot for SharePoint engine.
1. On the server that holds all SSAS engine (DBA team's server)
2. On the new dedicated SharePoint BI application server
Theoratically, PowerPivot for SharePoint engine is only used by SharePoint, so it is better not mixed with those SSAS instance that needs DBA to assign permission. Of couse it'd be better if we can have a dedicated server for PowerPivot engine,
but we don't. If it is mixed with MD Cube engine and Tabular Mode engine, it might compete resource more than put it on dedicated SharePoint BI application server. In addition, will separating PowerPivot service application with its engine cause more network
traffic and affect performance. Will it be better if we just put it on SP2013 dedicated BI Application server?
Like to hear your opinion.
Ming

Answer 1

Here's a way that I've seen implemented to switch sources amongst workbooks:
Create a "master file" Excel workbook that contains only the PowerPivot data model, and a blank Excel worksheet.  Upload this to SharePoint and create a BISM connection in the PowerPivot Gallery that points at this workbook url.  Here's a link
describing a BISM connection: 
http://technet.microsoft.com/en-us/library/gg471575.aspx
If SharePoint is configured correctly, you can click on the icon for an Excel workbook from the BISM connection in the gallery.  This will open a thin client Excel workbook that has a data connection to the BISM.
Develop your reports in the thin client Excel workbooks.  The reports will be smaller files since the entire data model is not in them.  Also, all reports will be pointing to the same data model hosted on SharePoint (via the BISM connection) to
keep things maintainable.
Treat the "master file" Excel workbook containing the PowerPivot data model in the same way that you would treat an OLAP cube.  Keep Dev, UAT, and Prod versions in each environment.
When migrating a new report from Dev to UAT, just modify the data connection to point at the UAT BISM instead of the Dev BISM.
Long term, if a decision is made to make the solution an SSAS Tabular instance, the BISM connection can be "re-pointed" at the SSAS instance and the reports will still work.  In this way, your reports can have Dev/UAT/Prod environments, the ability
to promote the solution to SSAS, and there should not be a need for rework if the design of the underlying model is maintained.