App-V - "Excel & ESSBase Plugin" perform poorly

Question

Hi,
using App-V 5.0 sp3 client for RDS 2012 R2.
*Excel 2007 & ESSBase Add-in sequenced into a connection Group
*Standalone Mode/ Normal mode so no streaming server 
*The App-V client UI (or PoSh commandlet) is showing package is fully loaded in local cache
*when I run Excel v-App, Excel Plugin "ESSBase" is registered fine
  I can open large .xls files with no issue
The performance issue escalated by a user is when loading a .xlsm calc sheet
and starting a data retrieval from ESSBase database server
- the operation takes around 2min 50
- compared to a "native setup" where Excel & ESSBase are installed locally, the same operation takes 30 seconds
TCPView.exe shows that starting the action throws tons of new network connections to the ESSBase database server
Procmon.exe filtered on Excel.exe shows :
- that compared to a native install, the App-V setup has tons more result="Path Not Found" or "Name Not Found" 
TaskMgr shows that on the App-V setup, Excel.exe enters "Not Responding" time window many times during data retrieval
=> on Native setup, Excel.exe don't enter "Not Responding" status
The tests were made on two identical RDS Session Host 2012 r2 VMs, 4 GB, AMD Quad Core, 1Gb Ethernet
with as single load the Excel test
=> I would be surprised if this is a ressource bottleneck - Excel.exe uses max 90 Mo.
I'll do benchmark tests using dummy network traffic tests using
measure-command powershell commandlet
from Inside App-V bubble & compare it to the results from outside the App-V bubble.
while I'm sure this shoundn't be a App-V limitation in itself, to be honest not knowing what to optimize
is frustrating.
Any help would be welcome,
Thanks.
MCTS Windows Server Virtualization, Configuration

Answer 1

Fixed !
now the App-V app (Excel & ESSBase plugin) data load from the database server
has simply the same performance as in the Native Installation Environment or is even a bit faster...
The ‘guilty’ was : “COM Mode” setting witch was set up on the App-V package’s dynamic configuration
XML file as in
Isolated Mode witch according to App-V: On COM Isolation and Interaction
meant that it is being Virtualized
On – w/ GUID Spoofing
I deleted the App-V package & set that bit of code inside of the XML like below then
 tested again – the result is now satisfying !
<COM
Mode="Off"><integratedcomattributes
inprocessenabled="false" outofprocessenabled="true"><IntegratedCOMAttributes
InProcessEnabled="false"
OutOfProcessEnabled="true"/>
</integratedcomattributes></COM>
I watched the below TechEd North America Event
Project Virtual Reality Check: Microsoft App-V 5.0 Performance, Tuning, and Optimization (App-V PTO)
it is briefly said that to troubleshoot execution performance issues,
- check unnecessary Extention Points wich could slow execution
- double check COM Mode setting as it could also horribly slow the App-V App user experience
See also the section Writing directly to Excel
where The scripting Guys explain two methods
& how COM is known to be horribly slow …
sample powershell code snippets from
PowerShell and Excel: Fast, Safe, and Reliable
to inject cells & generate and Excel file…
This was not easy to find as the majority of the documentation around ‘performance tweaks’ for App-V was leading to the Streaming or Package Publishing process optimization wich was not my case – I was more interested
about optimizing a process started during the package’s execution when it was already fully cached locally.
Thanks.
MCTS Windows Server Virtualization, Configuration