Even to write “give up” on the same line as “APM” (application performance management) feels prejudicial to me. Abandonment of the domain is what many readers will take away from Art Wittman‘s “What’s Killing APM“, though, and the question deserves to be faced squarely.
Wittman describes a world in which “cash-strapped IT teams … [a]re using APM … less and less. The … tools themselves … are notoriously hard to set up. . . . If you had a working APM … it has been broken …” by infrastructure changes. Fewer survey respondents than two years ago judge APM “crucially important”. Users have grown habituated to outages, so APM brings little gain.
I have methodological reservations about survey results. While I trust the integrity of InformationWeek procedures, a 3% slip in one particular aggregate score of two different samples of managers and executives rendering opinions separated by two years about what other people (the users of applications) are thinking sounds to me like a weak signal.
Rather than strain over the statistics of survey instruments, I prefer to chase down specific strategies and approaches that promote success. Body counts or popularity contests interest me far less than a winning formula for APM roll-out.
APM certainly faces plenty of challenges, including:
Is it time, then, to give up on APM?
For a serious answer, start with clarity about your own situation: is the end-user experience (EUE) that matters to you that of office employees directly connected to the corporate LAN, or of prospective customers reaching you through cellphones, or distribution-center workers operating through wireless tablets, or some combination of other audiences? How much does variance in performance manage to your end-users? Will those monitoring your APM dashboards have the authority and ability to take actions to correct problems? APM is utterly pointless for some organizations, and an absolute necessity for others. You’ll only know which yours is by specific and thoughtful analysis, not following the crowd which in 2012 might well be turning away from APM.
Recognize also that difficulty, in isolation, determines nothing about the right choice to make. Yes, it’s hard to do traditional APM across the boundaries of cloud provision. The lack of predictability in the cloud might mean, however, that APM becomes more essential as your operations move into the cloud. Business decisions must be founded on comparative payoffs and analysis of alternatives, rather than difficulty.
The right course, therefore, demands detailed planning and execution. Do you need APM? If so, what kind of APM? Which products have the technical ability to measure your application mix usefully? What will it take to acquire and train operators who know how to run APM?
APM is far from dead. It’s the closest we have to effective protection against outages like Amazon’s most recent storage lapse. Now is the time to choose the APM that best fits your own requirements, not to concede that application management is a mistake. It’s certainly not time to give up because of what others say.