Citrix XenApp Performance Essentials: A practical guide for tuning and optimizing the performance of XenApp farms using real-world examples

Health Monitoring & Recovery (HMR) is a tool included in Citrix XenApp which you can use to run scheduled tests on your servers to monitor their health, and if a problem is detected, you can automatically perform recovery actions.

It includes a standard set of tests, but you can import additional tests downloaded from the Internet or even develop custom ones.

When you install XenApp on a Windows server, the setup adds some new performance counters you can access from Windows Performance Monitor. A regular analysis of performance data helps to identify possible bottlenecks or lack of free resources.

Citrix performance counters are grouped into six sections. In the next section, you can find a description of the most significant ones.

If you're analysing the performance of your farm, in search of possible bottlenecks, you'll most likely find lack of CPU or memory resources.

Citrix CPU Utilization Management is a feature you can enable to improve the way XenApp manages CPU resources and to normalize CPU peaks.

When you enable CPU Utilization Management, the server checks the proportion of CPU resource (share) assigned to each user. By default, it assigns the same share to all the users. This prevents one user that executes CPU-intensive operations from impacting the other users on the same session-host servers.

For example, consider a user who is using a spreadsheet to perform complex financial calculations. Without CPU Utilization Management, his session will probably be consuming most of the CPU resources available on the server, slowing down the work of other connected users. With CPU Utilization...

Each session consumes part of the available memory on your server. If the memory request exceeds the available memory, the operating system needs to swap some of the RAM memory to disk, degrading the overall performance of the server.

The memory optimization feature can reduce the memory utilization improving DLL-sharing among applications.

Citrix XenApp runs on Windows 2008 R2. It's possible to apply some changes to the OS configuration to increase the overall performances of the server.

A document including all the suggested modifications has been published by Citrix (http://support.citrix.com/article/CTX131577). In the next section I'm listing, in my opinion, the most important ones.

Worker groups are a collection of XenApp servers.

A worker group can contain any number of servers from the same farm it belongs to. It can also be empty. A server can belong to different worker groups.

With worker groups you can:

You learned that during the resolution process, the data collector chooses the least loaded server to launch the requested application. How can it determine the load level for each server?

XenApp calculates the load on a server using load evaluators. Each server must have one and only one load evaluator assigned to it. You can also attach a load evaluator to a single application.

Each load evaluator contains one or more rules. A rule defines an operational range for the server or the published application. When any rule reports full load or exceeds its threshold, XenApp considers the server unavailable and removes it from the list of balanced servers. When all the rules return values below the configured thresholds, the server is added to the list again.

With load evaluators, you can change how XenApp evaluates the load of servers or applications. Load evaluator rules are based on performance indicators (CPU, memory, paging, and so on), not on the importance of your users.

Using Preferential Load Balancing, you can assign importance levels (Low, Normal, or High) to users and applications. Users or applications with higher importance level receive more computing resources.

Preferential Load Balancing calculates the resource allotment for each session, considering both user and application importance levels.

Enabling Preferential Load Balancing

To use Preferential Load Balancing, you first need to enable the feature with a server policy:

Enabling Preferential Load Balancing

Then, you can change your users' importance with the creation of different user policies:

Assigning High importance with user policy

Finally, you can change the importance...

In this chapter you learned how to monitor your XenApp infrastructure and the most important indicators that may warn you about overloaded servers.

With two XenApp features (CPU Utilization Management and Memory Optimization) you can optimize the usage of CPU and RAM, while changing some Windows settings you can optimize the OS for hosting ICA sessions.

Using Worker groups, Load evaluators and Preferential Load Balancing you can also improve how applications and sessions are balanced on your servers, with the ability to configure different importance levels for users and applications.