Remember when you were a child and your parents defined a set of rules that you had to follow, or chores you had to complete? You didn’t always follow all the rules or complete every task, but you paid close attention to what your parents really cared about. As adults, we still don’t like to do everything we are told, and we prioritize our actions based on what we (or our superiors) care about. If it isn’t important to anyone, then why do it?
The same principle is true in IT. Many elements of the infrastructure can be monitored, but if no one cares about the end result then it’s ignored, and is nothing more than noise. This is why when monitoring the IT infrastructure, teams need to consider what specific information is crucial to capture, who needs to know the information, and what he or she will do with it. If an alert is raised, who on the IT team needs to be notified? Is it the SCOM operations staff or is it a specific domain administrator?
In order to truly change IT behavior, metrics need to have two qualities:
- They must be actionable: Monitoring metrics are meaningless if there’s no associated action that comes as a result. For example, if you’re monitoring memory or CPU in seconds, it doesn’t mean much, but if you use a percentage or a rate to quantify it, it can provide value and an actionable response. When monitoring the IT stack, you must ask the question, “Who will care if this alert is generated or if this metric is graphed?” Bottom line: it must be relevant to someone or it is a moot – and ultimately wasteful – exercise.
- They must be mission critical: But it goes a step further than just needing to be actionable. There are many things that are monitored that are actionable, but where it really becomes a behavior change is if the monitored item has an impact on critical business applications. To know if this is the case, teams must know which applications are associated with which alerts, and who will care about the alert. If a Load Balancer Virtual Server is down for an application that is depreciated, no one cares. But if Microsoft Exchange is down, everyone cares. In essence, the criticality of the application is what will impact behavior.
If you’re not monitoring the infrastructure with the goal of keeping business critical applications running, then your monitoring has little value and is just a mundane chore. To ensure our customers execute comprehensive monitoring, we identify the performance characteristics of each virtual server in the environment in our F5 BIG-IP management pack for SCOM. In addition, our Nutanix management pack has an industry-leading application awareness feature that will identify if a virtual machine (VM) is running a Citrix Delivery Controller, StoreFront, or a variety of other Citrix roles (additional applications will be discovered in upcoming releases). This will help the SCOM operations staff and the Nutanix administrator determine if anyone really cares if a VM is overloaded or is having a high IO latency.
So what’s the big takeaway? If an alert is actionable and impacts a critical business application, you care.