“Performance” means something new in a rationalized cloud architecture

Are your operations mediocre? By the definition of mediocre, most are. If so, you have a greater need of application performance management (APM) than if your IT (information technology) shop were already finely tuned.

To understand why is rather complex. Consider, for a moment, a much different subject: Sarbanes-Oxley (Sarbox) compliance. Managerially, it’s easy to explain: the US Congress passed a law in 2002 which requires IT to do and not do certain things. While the details are complex enough to spawn a whole industry, the basic business imperative is perfectly clear.

APM is almost the exact opposite: however straightforward it sounds superficially, its true meaning and consequences require technical depth to capture. APM measures response times of applications, raises alarms when they go “out of range”, and at least begins investigation into causes of problems.

The first generation of APM focused on “in-home” software; it assumed that source code was available for all deployables, and often further assumed that the source was in one of only a handful of languages.

APM’s current responsibilities are so different that the marketplace is almost unrecognizable from a perspective formed just a few years ago. Modern APM needs to monitor all mission-critical applications, whether developed on-site, licensed from a vendor, installed from open source, or any combination. It needs to handle not just a basic client-server architecture, but any deployment model, including ones buttressed by reliance on third-party services, public or private cloud hosting, distributed infrastructures, and virtualized everything.

And APM matters now in a way it didn’t just a few years ago. Then, traditional datacenters were filled with over-scoped servers. When each application had its own physical server, it was customary to provision the latter generously. With typical server utilization of 10-40%, performance monitoring had few consequences. Servers were over-scoped enough that they effectively handled most spikes.

Now, though, with on-demand scaling and executive-level imperatives to quantify efficiency, reserves are vanishing. A hiccough in the network-attached storage for a reverse proxy might end up severely degrading query results in an apparently unrelated database management system.

That’s far from the end of the story, too. A strategic analysis of APM applicability from even a couple of years ago deserves a complete rewrite now, as trends in datacenter design, software-defined operations, utility computing, cloud, mobile, and more all slice up our operational stacks and re-combine them in new, hard-to-manage ways. You might have looked at APM in 2010, and concluded you could easily live without it; figure its return-on-investment now, though, especially if your operations have always been relatively informal and relaxed, and you’re likely to find APM is the only way you can stay on top of all the expectations for IT.