Application Performance Management (APM) is a hot topic right now. Gartner defines APM as agent-based monitoring that sits inside the operating system and provides code-level performance, tracing, application mapping, and tracking. How exactly does APM help an organization, and when would a business choose to invest in this technology? When does APM make sense and when doesn’t it? And, more broadly, how does this tie into the changing needs of IT monitoring? Finally, why does the agent vs. agentless debate continue to rage on?

Simply put, enterprises that write their own code (Java, .NET, etc.) and leverage applications unique to the way they do business must have code-level application visibility. More specifically, those companies who place high importance on understanding how code executes and functions in a production environment, and what that means to business-critical, revenue generating, bespoke applications need APM.

That said, APM is not necessary for the vast majority of commercial applications not authored by the enterprise because code-level visibility is not necessary, for instance in the example of a CAD app purchased from a provider of an ERP solution. There is also the cost consideration. As a single APM agent typically runs somewhere between $150-$200 per month, from a cost perspective it simply doesn’t make sense. If your authentication service goes down, you’re not going to use an APM agent on that. In fact, most of your operators wouldn’t even know what to do with the deep code level data you’re getting back.

Today we’re seeing traditional IT infrastructure management vendors moving towards an application-centric view of the world and APM vendors attempting to get broader visibility of the entire IT infrastructure. As an enterprise, I need to understand how all of my infrastructure is working — what’s up, what’s down, what’s running well and what’s not, capacity planning, failure analysis, and keeping the lights on across my vast complicated set of IT technologies. Simultaneously, organizations need to know how their applications are doing. However, rather than handpicking one or two “important” ones for code level visibility, you’d really like the two different types of vendors to meet in the middle.

So most organizations are combining application-aware infrastructure monitoring for all apps and augmenting in spot places with APM for custom apps.

On to the war — agent-based versus agentless monitoring. For years now we’ve heard sniping back and forth as to which model is best suited for enterprise IT. Both approaches have their pros and cons. Agents can provide more granular performance metrics, while agentless monitoring platforms are often easier to manage. But to say you can only have one or the other is a canard. There are vendors that provide customers with the option to deploy both models simultaneously, depending on the customer’s need.

If there is one inalienable truth concerning IT, it’s that IT has and always will be heterogeneous in nature. The complexity of systems and IT infrastructure ecosystems demand it and IT will never converge on homogeneity. Enterprise IT should not look to choose between APM and application-aware infrastructure monitoring. Nor should they be forced to adopt a single approach to gathering performance metrics. That of course isn’t stopping vendors from yelling from the rooftops.