In my last post, we started to talk about a performance management roadmap. In this post, I’d like to share some of my thoughts around the nature of the APM problem that we are trying to solve and the economic implications of that.
We’ve been working around this theory that the goal of APM is to identify the root cause of any issue whether it has happened, is happening, or will happen, and then take protective measures to prevent it from repeating. One approach to solving this problem says that it’s key to have every possible metric generated by an environment collected and kept historically. While this is impossible from a number of dimensions, let’s exercise this philosophy a bit. Disregarding the frequency of collection, what are the implications of capturing every metric that is potentially capable of being captured?
See the following graph.
The graphs and data represented here are not based upon any independent source of knowledge. They are all relative and purely to establish a basis of conversation. In the graphs, “time” is representative of modern IT, dating back to the 1960s and moving through the next decade. If we were to assess the total number of available metrics to be measured – it’s reasonable to assume that global number of IT metrics to capture has been increasing, at an increasing rate. This is essentially an exponential curve of some degree. It’s possible that that curve is higher or lower in pitch than the one represented in the upper left graph. This belief is based upon growth of open systems, emergence of desktop computing, industrialization of India, China and other global regions, mobilization of the workforce, and the development of the Internet industry. Each of these macro waves have acted as a catalyst for the expansion of computing systems, increasing number of technological layers in the data center (web servers, networking, storage, app infrastructure, databases), and increasing numbers of end users.
At the same time, the cost to manage a metric started low due to the simplification, out-of-the box, holistic management offered by mainframes. Those solutions shipped with the management solutions pre-integrated. As the industry moved to open systems and variety exploded, so did the challenge and cost of managing a single metric. Different formats, standards, environments all created challenges in understanding the metrics, gaining access to them, normalizing them, and organizing them in an appropriate way. So, the cost to manage a single metric increased with the emergence of open systems. And, over the past couple decades, as we’ve become knowledgeable in event, log, and performance management, we’ve slowly been making it cheaper to manage each metric. Standards like SNMP, Dell OpenManage, JMX, WMI, WS-Management are simplifying what it takes to gain access to and manage any single metric in an environment.
The problem is what happens when the total number of metrics to be managed and the cost of managing a metric are combined. While – as an industry – we are making it cheaper to manage single metrics, the volume of metrics to manage have grown so fast, that the total cost to manage all of those metrics is also an exponential curve. Given IT budgets are flat to slight growth year over year, it’s not going to be economically feasible for organizations to manage everything that needs managing. They are going to have to make trade offs – and open up the risk of not being able to predict or source the root cause of an issue because of gaps in coverage.
According to Gartner and Forrester, the ECA, SLM, BSM, and performance management industries are growing anywhere from 15-35% compounded a year (depends upon the source and the orientation taken to the data). This is faster growth than the single digit growth seen in IT budgets, which means that performance management technologies are taking an increasingly larger share of the IT budget each year. That’s CFO for, “Achieving SLA compliance is getting more important.” But even at 35% growth on the $2B industry, IT organizations will not be able to keep pace with the total volume of metrics that they have to manage.
So, what’s the resolution? The vendors have to DRAMATICALLY change the direction of the cost curve. A linear improvement in making it cheaper to manage an individual metric isn’t good enough. We have to innovate in ways that we drop the cost to manage an individual metric by 90-99% in order to make it economically feasible for organizations to implement strategies to get total metric coverage, eliminate all gaps, and ultimately achieve the goal of sourcing the root cause of any issue.
But how? What will it take for vendors to disrupt the cost curve and to put it on a different trajectory? We’ve got a lot of ideas about that at Quest. I’ll be writing about the different technologies and philosophies we are researching, exploring, implementing and delivering in upcoming posts.
Tyler Jewell
Sr. Dir, PM
Quest