Julie Craig is right: “… companies are monitoring and managing complex interations … at a very superficial level.” As the Research Director for Application Management at Enterprise Management Associates recognizes, though, this is an unprecedentedly challenging time for Application Performance Management (APM). Her article, “What Do Cloud, Integration, Mobile and Big Data Have in Common?“, underlines the extent to which APM vendors and users have fallen behind the technologies they’re supposed to control.
It’s not particularly that anyone has done anything wrong, or even that we don’t know how to do APM right. It’s simply that deployment practices are changing too rapidly, and APM hasn’t been a priority in the way time-to-market, scalability, end-user experience, and even availability are for most organizations. Practice and toolsets in APM haven’t had adequate opportunity to catch up with the changes brought on by Cloud, Mobile, Big Data, and Integration.
The APM challenge is even greater, in fact, than Craig sketches. She correctly identifies “… Cloud, Integration, Mobile, and Big Data … [as] four of today’s hottest technologies”, and certainly the ones likeliest to sound familiar to non-technical business decision-makers. Two other “megatrends” deserve similar analysis: adoption of software-defined networking (SDN), and content-delivery network as-a-service (CDNaaS). SDN and CDNaaS don’t make it into the trade press the way “Mobile” and “Cloud” do, but they have the potential to revolutionize application “plumbing” as much as the other four technologies are altering application implementation.
CDNaaS, as provided by Amazon CloudFront and others, might be regarded as a species of Cloud. Perhaps APM will treat it that way some day, alongside infrastructure-as-a-service (IaaS), platform-as-a-service (PaaS), software-as-a-service (Saas), and other Cloud offerings. For now, though, CDN has a complicated relationship to APM that, I think, merits separate consideration. CDN providers have generally been leaders in deployment of APM internally. At the same time, CDN is often acquired late in development, as a remedy for problems uncovered by APM. To the best of my knowledge, there is not yet a model for CDN or CDNaaS exposure of APM that integrates well the monitoring an organization needs for the other technologies Craig lists. I repeat: in principle, the CDN marketplace can provide all of Gartner’s Five Dimensions of APM. It just hasn’t happened yet.
SDN presents, if anything, even greater APM mysteries. As this year’s USENIX Workshop on Hot Topics headlined, “Little is known about the practical performance implications of Software-Defined Networking …” What we do know is that “Industry is embracing SDN”, and that APM is far from mature in this area.
Finally, as Craig touches obliquely in her comments on Integration, APM as a discipline still lacks polished solutions for components based on neither JEE nor .NET.
The conclusion? This is a great time for engineering and business leadership in APM. Perfected solutions are simply not available at any price yet, so devops teams need to analyze their own situations carefully, estimate opportunities and risks, and choose partial answers based on incomplete information. We’re still in the “craft” stage of APM, where considerable customization and “fitting” is warranted. If done with care, APM will bring all the advantages of bandwidth utilization, lowered cost, enhanced availability and responsiveness, and improved manageability.