The quote “Everything Old is New Again” can be attributed to lots of people. Performer Peter Allen penned a song by that name. So have the Barenaked Ladies. A character in Stephen King’s 2005 novel “The Colorado Kid” also utters the phrase. I, for one, hope there’s some truth to it. After all, that would mean it’s just a matter of time before I can grow back my sideburns and mustache and my closet full of pastel-colored linen blazers (ala Miami Vice) will again see the light of day, right? This quote also comes to mind for the topic of VDI or Virtual Desktop Infrastructure too.
For the uninitiated, Virtual Desktop Infrastructure (VDI) is virtualization technology that hosts a desktop operating system on a centralized server in a data center. This OS is then provided to end-users on demand. Sure, it sounds new and sexy, but when you get down to the brass tacks, the concept isn’t that much different from what IBM was doing with mainframes and dumb terminals in 1965. Regardless, just like bell-bottoms and wooden clogs in the 1970s, VDI is all the rage today in corporate IT. Therefore, it needs to be on the radar of the people responsible for monitoring and managing technical resources.
“On the radar” is the operative phrase here. It is incumbent upon us as IT pros to be aware of new (and old) technologies. However, as mentioned in this blog in months past, network and systems management is as much art as it is science. Sometimes the trick is knowing what NOT to monitor. Hopefully, the irony of an NMS vendor recommending you NOT monitor something doesn’t go unnoticed, but that is exactly what best practices dictate for a variety of reasons.
The first reason it isn’t particularly wise to monitor Virtual Desktop Infrastructure instances is the same reason it isn’t wise to monitor physical workstations and desktops: They are typically endpoint nodes in a networks and consumers of resources rather than providers of services. As such desktops/workstations tend to come up in the morning and go down in the evening when employees leave. Why monitor devices and alert on devices you know are going to be down everyday?
The daily UP and DOWN states are just one aspect of this issue. Another often-overlooked problem is licensing costs. Most active-monitoring NMS systems today base their license consumption on one of two things: “devices” or ‘polled interfaces”. Your typical IT shop will most definitely reap efficiency gains by deploying VDI in place of physical desktop computers. However, astronomical licensing fees can easily nullify those gains. It is far better to spend a tools budget on getting visibility to IT gear crucial to the day-to-day operation of the business rather than watching devices that are likely down 14 – 16 hours a day.
Of course monitoring UP and DOWN states is only one part of what’s normally monitored on computer systems (as differentiated from networking gear). NMS tools typically also watch variables like CPU utilization per process, disk usage, running services, and others. However, the magic of VDI is that once a desktop is no longer in use its resources are relinquished to the issuing hypervisor for allocation elsewhere. In other words, VDI desktops are completely dynamic. If you’re monitoring a VDI instance for a given problem and then the user shuts down and goes home what happens to all that historical data already gathered? Poof! It disappears faster than face paint and spandex at Comic Con. The transient nature of virtual desktops makes them impractical to monitor.
None of what’s been written so far is to say that no VDI monitoring is the best course of action. There are aspects of the technology that make perfect sense to keep track of from a performance and availability perspective. For example, in a large VDI deployment it is very likely that all desktops aren’t getting served from a single hypervisor, but instead a load-balanced cluster. The health of all the hosts in that cluster should be monitored for optimal health. The same is true for the performance and operational status of any data stores in that VDI cluster (for example disk I/O, kernel latency, and their ilk).
Given the abundance of free and commercial NMS tools available for IT managers today there really is no reasonable excuse to NOT monitor your systems and network infrastructure. Just make sure you’re picking the right equipment to monitor. Unlike the parachute pants and denim jacket that I’m planning on wearing into the office for Casual Friday that’s an idea that will never go out of style.
