Know why you’re doing it. Analytics are admittedly confusing. For instance, anomaly detection is central to security and performance management, but can also be leveraged by more advanced functions in configuration and change management, or compliance audits among other areas.
While you may well seek to optimize anomaly-capable analytics be wary of trying to stretch a single solution too far. At the same time, look for redundancies. You may find that a super solution for security carries over well into some areas of performance or vice versa. This is rare but true in some cases. When this kind of “collapsing” is possible it not only saves money, it can help to build bridges between teams that usually don’t communicate.
Model-enriched analytics, as it’s evolving, is an especially good investment. In fact you’ll see examples from virtually all the platform vendors including exciting new announcements from several of the major ones this quarter. The reason for this is that modeling — especially as an extension of a core CMS foundation — links physical and logical interdependencies like topological, application dependency, or configuration histories on the physical side with “owner,” “customer,” and “business outcome,” or “business process,” on the logical front.
Some modeling can also help to prioritize trusted sources so that redundant key performance indicators (KPI) coming in from multiple management tools don’t flood your analytics with disruptive inputs. And so modeling can drastically enhance the value of analytic investments, especially those analytics that assimilate information from many different sources.
Of course, to do this the modeling has to be dynamic and current. Once a week or even 24 hour updates won’t cut it. And admittedly, while such dynamic modeling capabilities are currently available in the market it is still early in the game. Still and all, model-enriched analytics is, in my opinion, the space to watch when it comes to the future of analytics and to a large degree service management in general.
My last bit of advice is to use common sense in the face of dazzling claims and rocket-science-driven arguments for supremacy. All of the more advanced analytic capabilities minimize deployment and administrative overhead; usually quite dramatically. This is true in spite of the fact that some vendor claims for “no administration” are a tad inflated, as someone has to mind the store in terms of which KPIs to assimilate and how the analytics need to be integrated into the bigger picture. But if you see a deployment that requires a lot of resources on your part look somewhere else! There are too many really good options out there in 2011 to waste your time with analytic technologies that seem to require an infinite number of Marines. Your biggest challenges may once again be cultural: having the organizational structure and processes in place to actually optimize such a cohesive and dynamic system.