CPU and memory issues on your host servers can adversely affect the end user experience. As with most technical solutions, we recommend that, while you are going through your initial testing and rolling out, you document your baseline for key performance characteristics, such as CPU and memory utilization, and deeper metrics such as CPU-ready times.

With this baseline information in your toolkit, it is easier to compare the performance before the issue occurred to what you are seeing now. This will help to try and pinpoint what could be causing the problem. Likewise, using technology such as vRealize Operations for Horizon will help you understand performance utilization over time. You could use a third-party product such as Liquidware Stratusphere. Stratusphere provides the baseline metrics (measured during the assessment phase), as well as the current user experience, allowing you to quickly identify what changed, and what is causing the issue.

Within your VDI infrastructure, you don't want to be experiencing any memory overcommits. You need to consider how much memory is allocated to your virtual desktops, as well as the total memory within your hosts. Ideally, you want to ensure that your total allocated memory is less than the total in your host servers. Don't forget that should one of the host servers fail, you will still need to supply the required memory resource.

If you are experiencing performance issues related to memory or CPU, check whether memory is being swapped by any of the virtual desktop machines. Check whether there is any ballooning within the environment. Understand what your CPU-ready characteristics are. The acceptable CPU-ready figures within your VDI environment will vary based on the environment and users. Generally speaking, you are going to want to keep CPU-ready below 5% per allocated CPU, with 10% at peak. Anything over these numbers would indicate that the host server is being overloaded, and will impact the performance of the CPU and lead to slow performance of the virtual desktop machines. The following screenshot shows an example of CPU monitoring and, in particular, the CPU-ready time:

It can be very easy when growing your VDI solution from the initial design to forget to sanity-check these metrics and keep an eye on them as you grow, and, all of a sudden, you may find you have a compute performance issue.