This is part two of my earlier blog Seek and de…eh..load test the VDI – Login VSI – part one what and how? Part one can be found here at this post This part will be about the test driving in my test lab and a little conclusion. A notice up front; my testlab is not capable for a oh how screaming fast can this VD be (cause it won’t be), this is just for test driving the configuration and options to get a technical feel about the Login VSI product. But first some recap from part one.

A little recap

In part one I explained the need of testing and how Login VSI can be of great help here. I explained the components needed in a Login VSI, the architecture model from part one to help.


Using Login VSI needs a:

– VSI server with file share and web service for launcher and targets to connect to.
– One or more launchers. The launchers need client components, a RDP client for RDP sessions or VMware Horizon client when connecting to a Horizon VD target pool.
– Targets. VD environment for running the workload.
– AD DS directory for users, groups, OU and GPO to help with the workload setup.


Testdriving will start where we left off in part one. From part one we know and got a data server, a launcher and target environment. Setup the hostnames (even the HTTP URL) with short names. 

I have setup the target machines in a standard configuration, base Windows 7 image, member of domain, office 2013, View Agent and ran the target setup. The target setup adds the remaining applications to the image and add the VSI security/user group to the RDP members.  


I then used the VMware OS Customization Tool fling wizard to optimize the OS somewhat (you can find this great fling at Created a snap and added to a floating pool. I added the VSI group as entitled users. Important but easily overlooked, but this group has the Login VSI users in there.

Next up is the workload settings. Be sure to set it up correctly. I had to change the home / group drive letters as I had a personal disk that is mapped to the user home directory from Login VSI (and this will fail the tests). Fill in the language and Office version (Office 2013 for me).


Skip the debugging entries for a normal test.

I have done launchers on Windows 7 and Windows 2012, both work correctly for RDP target sessions (use short names!).  I setup the RDP connection to the VD first to check if standard connection is working. In another run I checked the VMware Horizon view client, therefore I installed the Windows view client on the launchers (from the view server root page (https://view) and link to client to be redirected to

On the launchers be sure to start the launcher/agent in _VSI_Binaries before starting the tests. The VSI data server will need a started session monitor. They can be combined on the same server (for small environment that is), just setup correctly and run the session monitor and agent before the tests.


A RDP Agent session.

Next start up the Management Console to get ready to rumble. Check all servers setup correctly, UNC paths filled in, change the web service url to HTTP://shortname.

Choose or define your workload scenario. Pro edition will give you access to the pro content. This is an additional 12 GB worth of testing. This library contains extra data files which are used by the heavy workload. The idea is to have a ‘larger’ pool of ‘random’ files (think of caching) that will be processed during a workload. Optionally add some phases to simulate a changing workload.

And create a connection at test setup. For RDP use the RDP wizard, for view use the VMware view wizard. User is LoginVSI{count} to increase to users to the amount you specified in the test. The desktop name in the view setup is actually your pool ID. A note for testlabs, when using self signed certificates be sure to import them in the trusted root certificate store, and be sure to license your view testlab. Else no connection will run (you will have to switch from non interactive mode to check via command prompt).


Finally use start test to setup the test. The test name that is filled in is also the name at which it will be save in the vsishare (logs results and such) choose an appropriate name for reference. Use a workflow user (only with Pro) or use the standard users as created in the setup.

Important before your first launch is to suppress the standard setup profile waiting processes be sure to yes, no, trust continue, setups et al. Or login with your workload users to set this manually; and yes this is only workable for a small amount of users.

Launcher check will be okay (if you started agent of course). On the agent command prompt you will see that it has informed the launcher. Finishing will start the monitor and workloads are started on the target.


When you wish you can monitor the workloads on the launcher. You will see a session being opened and workloads to be started. Start your first test with a small amount of sessions to see if everything is setup correctly (yes this is the testing part) and monitor these sessions. You will see errors in the sessions if something is stuck, not configured correctly or not responding.



After the run you can open the analyzer for those graphs, this is the part you where waiting for. Open the _VSI_Logfiles/name as you named the test in start test.



When I was testing a new version 4.10 was beta released. Didn’t find in the release notes if this is a preferred clean install (as from v3 to v4), but from the type of update this can easily be a in place. You can do a clean one. Save your tests for further reference before deleting the old installation. As I was testlabbing (can this be a new dictonary word?) I have done a inplace upgrade and that works fine.

Great additions are the type of workers scenario’s, one for the knowledge worker and one for the task worker. Also nice is the beta import from external sources like perfmon and esxtop sources. I haven’t tried those two. I noticed that the VMware Horizon client is also in the right folder.


There are two Login VSI editions. One is the express version aka the trail version. Limited to 50 user workload, 30 day evaluation time of usage and limited in functionality (vs the Pro). And there is the Pro license version. Like written before this give you access to pro library for extra workload items. And this will give unlimted functions and amount of users you bought a license for. This following table compares the two versions.



On first glance it is not as easy straightforward as claimed. It is not ready to run from the box. You need some experience with this product to install correctly, know all options and components and how they interact. And how to interpret the values…..
Get to know the scenario’s and figure out how to make conclusions based on the received outputs and your environment. With some testing the of the scenario’s you will get the hang of it in no time. Read the manual and don’t be stubborn like me if the manual says use short names, use short names and no fully qualified.

This product definitely adds to the VD project consultant and engineers toolkit. For heterogeneous consultants working on different sorts of VD implementation this is the standard workload load testing tool to strap to your Batman tool belt. 
Sure there is a small investment (commercial max number of user workload license) but it earns back (and even more) at the start by giving valuable insights and reporting to your projects. This lowers project execution time, project budget and mitigates the risks of non performance to your production environment, non existing user experience and disgruntled users. This in turn lowers the possible costs of rework of these non performing incorrect configurations.

I would like to see an installer wizard we there is an option to install and configure a default VSI server, launcher with the options set to a defaulted out of the box working installation. This makes it even more straightforward and gives you to possibility to learn the product before changing the setup for larger testing environments. And this increases portability for the analyzing consultants.
Next on the wishlist is a location to upload your results and check with other users result sets. You then have something to compare to around the field. Now you rely on the experience and workload knowledge of the consultant doing the test.


Maybe I will need to fit in a part three about the analyzer graphs….hmm… ponders………..

– Happy load testing!

Share this: