[GIS] Performance testing ArcGIS Desktop in Citrix environment

arcgis-10.0arcgis-desktopperformance

I have a project (ArcGIS 10.0) for a client that could potentially be deployed in a Citrix environment. It will contain a fairly complex Add-In that will, among other things, import data, allow them to perform reclassifications on rasters, digitize features, and ultimately let them setup, through the Add-In interface, parameters for a 3D modeling scenario in a package outside of ArcMap. These setup parameters will probably be written to a XML file via .NET in the Add-In. The application of course isn't developed yet (as we are in the early stages of the project), but we need to test as much of the future functionality of the application as we can in their Citrix environment and on their local PC installs as well. We have test data to use.

How do I go about performing tests that can provide metrics to compare performance of ArcGIS in their Citrix environment versus their local ArcGIS installs – are the there any software packages and/or methods/protocols I can employ?


We have thought about this some more and are thinking of doing some qualitative tests such as having bookmarks in the ArcMap project and having the user across the country time the draw speed, see how editing of point and line features performs, and 3D display in ArcScene. These are obviously very qualitative and subjective, but it's one approach.

Best Answer

For what it's worth, here is what we ended up doing. We created an ArcMap project, added some large-coverage raster and vector datasets to it and created a few bookmarks in it. We then wrote up a "recipe" for our remote user to walk through - things such as:

  • Zooming out to full extent with certain layers on/off
  • Moving through a list of bookmarks with certain layers on/off
  • Editing vector line and point featureclass in a file geodatabase
  • Zooming and panning randomly at will

The MXD and file geodatabase feeding it were both located in the southeastern U.S. (where we were at the time) and our remote user was located in southern California. We had him measure how long most of the recipe steps took, and he also gave comments such as "that was almost painfully slow" (really). We knew going in that these would be mostly qualitative metrics, and we were okay with that. Turns out the performance, for the most part, wasn't too bad. Worst part was drawing large rasters and large coverage polyline layers with tons of vertices (what we expected). We then performed the same recipe steps on ArcMap in Citrix from where we were located (with the Citrix server right down the street from us as opposed to across the country) and it worked fine in that situation.