LoginVSI, Session Metrics, FSlogix Cloud Cache and some performance counters

Recently I got involved in multiple projects implementing and resolving issues relating FSlogix.

At the same time I noticed that as of Login Enterprise 5.4 session metrics can be displayed in load and continuous tests.

This doesn’t sound big but is actually filling a big gap for IT consultants using this tool! With session metrics you can capture performance counters from within the session during the test and take that with you in the test results without an additional agent or tool. The most obvious of course are CPU and Memory usage but let’s play around.. what else can we do?

Things like VMware Horizon Blast performance, Citrix PVS writecache usage (always fun) and in this blog post I will capture FSlogix Cloud Cache behavior!

First

let us get the obvious out of the way, 3 must reads:

1. What is FSlogix Cloud cache: FSLogix and using the “OnPrem” Cloud cache (debruinonline.net)

2. What are Login VSI Session Metrics: Session Metrics – Login VSI

3. How to use the Login VSI API, used to create your own counters: Getting Started with Login Enterprise v4 API - LoginVSI

The performance counters

So now we are all up to speed. We need to determine which performance counters to capture. To see the ones available, the most easiest way is to open perfmon on a VDI and add a counter or by using typeperf -q (see the sidetrack below.)

the sidetrack: see one of my older blogs to see the Citrix PVS performance counters Citrix Provisioning, Windows Performance Counters, Telegraf, InfluxDB and generate awesome graphs (debruinonline.net). Scroll down to the PVS target devices to get the Writecache usage. Really nice to have this data in a load an continuous test!

Back on track: Load testing VMware Horizon and FSLogix CCD behavior.

When scrolling down I found the FSlogix CCD Driver Performance Metric.

2 Counters go my attention. External Bytes Read per Second and Local Bytes Read per Second.

So how many bytes per second are gathered from the remote container and how many bytes per second are gathered from the local cache. Interesting!

Create the Session Metrics:

Define the External Bytes Read Per Second counter, mind that both the counters don’t have an instance (unlike the example by Login VSI) so you can leave that optional parameter out :

{
 "type": "PerformanceCounterDefinition",
 "counterCategory": "FSlogix CCD Driver Performance Metrics", 
 "counterName": "External Bytes Read Per Second", 
 "tag": null,
 "displayName": "External Bytes Read Per Second",
 "unit": "per seconds" 
}

Define the Local Bytes Read Per Second counter:

{
 "type": "PerformanceCounterDefinition",
 "counterCategory": "FSlogix CCD Driver Performance Metrics", 
 "counterName": "Local Bytes Read Per Second", 
 "tag": null,
 "displayName": "Local Bytes Read Per Second",
 "unit": "per seconds" 
}

And define the Metric group:

{
 "name": "FSlogix Cloudcache",
 "definitionKeys": [
   "ec8e72c9-e89b-4fb6-8968-f73cb65df8a8",
   "83bd5a87-7b20-430c-86b4-ca099372dbac",
  ]
}

The value of the definition keys are the ID’s you retrieve when you do the post of the Custom Metric. (you did read the “must read blog number 2” right?)

Added the new session Metric to my load test:

The tests:

I ran 2 tests. All on a non persistent Horizon VDI Instant Clone Pool so clean VDI’s every start.

All the users run the default Knowledge Worker 2022 Application group test (containing Microsoft 365 apps and actions)

Test 1: new VSI users and they don’t have any FSLogix containers yet.

As expected there are almost no External bytes to be read… since there are no external bytes . The bytes are read locally. Completely makes sense. The profile and office containers are cached locally and hydrated to the central file server, al the data is available locally.

Then I ran the test again, this time there are profile and office containers available at the central file server.

Test 2: this time there are profile and office containers available at the central file server. .

What you can see now is the data is Read externally and pulled down to the local cache. Further on during the test all the data is read from the local cache, there are almost no External reads later on.

More easy to see of you compare the results of the two tests in one graph:

2 other results worth mentioning, the Prepare for Microsoft Office 365 and start Microsoft Outlook Result. The power of load testing and comparing them with VSI starts shining here! See for yourself:

Conclusion:

This is a great addition to the tool! But in all honesty as a visualization junky I will still grab to additional tooling to see more details and layer these over one another.

But that is just me, great job Login VSI!

Need some help with this new feature or in generic with Login VSI, FSlogix, Desktop virtualization, or performance testing and visualization in generic?

Don't hesitate to contact me!

For Login VSI you can also join the growing Login VSI slack channel:

https://join.slack.com/t/lvsi-customers/shared_invite/zt-25er8f8ms-g4NFkAosV9MqTRRSiERGNQ