When an organization makes the transition from physical to virtual desktops, the IT department must do everything that they can to make sure that the virtual desktops perform as well, if not better than the physical desktops that they are replacing. To do so, administrators have long relied on performance metrics. Today however, there is a fundamental shift that is occurring with regard to the way that virtual desktops are monitored. To put it simply, the industry is finally beginning to accept the idea that the end-user experience is more important than raw performance metrics.
Performance metrics will always have their place, but raw metrics are open to interpretation. After all, there are sometimes situations in which users complain of slow performance even though the performance metrics fall within a healthy range. Conversely, user sessions might sometimes be performing really well even though the performance metrics seem to indicate that they shouldn’t be. When either of these situations occur it is a clear indication that you may not be monitoring the correct metrics.
If you really want to get the most out of your VDI monitoring software then you must figure out how to hone in on the end-user experience rather than merely examining a series of raw performance metrics. This requires doing three things:
The first step in the process is to figure out what types of activities the end users are most likely to use to gauge system performance. Generally speaking, users tend to equate overall performance with system responsiveness. In other words, the user’s perception of system performance is based largely on what the user is seeing or feeling. The main VDI UX metrics include:
When a shortcoming with the end-user experience is identified, the next step in the process is to identify the conditions that might be attributed to the poor level of performance that the user is receiving. Ultimately, user reports of slow virtual desktops can often be traced to either excessive resource consumption or to insufficient resource allocation. In either case, the key to resolving the issue is to identify the resource that is in short supply and then make any necessary adjustments. It is important to keep in mind however, that performance issues are not always hardware related. An improper OS configuration can just as easily cause performance problems. For example, if a configuration error causes a user to be authenticated by a remote domain controller then the user will experience a slower than normal logon even if the proper hardware has been allocated to the virtual desktop.
Some of the more commonly reported end-user experience problems and their most common causes include:
There are two main approaches used to measure end user experience metrics in VDI environments:
Each vendor has its own approach to synthetic transactions, but generally speaking the technology works by using simulated users to launch real sessions and then measure key performance metrics such as logon duration and app load times. Here are some vendors that use the Synthetic Transactions approach for VDI User-Experience monitoring:
The other approach that is commonly used is Real User Monitor, or RUM as it is sometimes referred to. As the name implies, Real User Monitoring works by monitoring real user sessions and keeping track of key performance metrics. Again, each vendor has its own way of doing things. Here are some vendors that use the RUM approach for VDI User-Experience monitoring:
Although it is easy to become preoccupied with virtual desktop performance metrics, it is ultimately the end user experience that really matters. The key to providing your users with a good experience is to understand the relationship between hardware resources, user activities, and performance metrics and then use that information in a way that allocates the necessary hardware resources to the individual virtual desktops. VDI admins should look for a VDI monitoring tool that monitors the user experience metrics that are most likely to impact the end user experience, such as logon duration, application load time, etc.