As foreseen in previous years, the Application Performance Management (APM) market continues to develop, both in terms of numbers of providers and inherent functionality, particularly with regard to end user visibility. However, a lack of fundamental differentiation between many of the providers means that consolidation is to be expected, either due to the collapse of (over geared) Vendors, or via acquisition.
New adopters of APM would be well advised to consider whether their favored vendor is likely to be a purchaser (and therefore provide continuity), or a purchase (and run the risk of absorption into a larger system, or disappearance). This advice would tend to favor the larger / more established Vendors. Partly, however, this will depend on the investment and time horizons for the chosen product. "An APM is for life, not just for Christmas" does not necessarily have to be a given, particularly for smaller companies with straightforward delivery infrastructure and little legacy baggage.
It is increasingly necessary to trade off the ongoing, expensive, skills requirements necessary to gain the most from some APMs versus the immediate (but perhaps ultimately more limited) value of competing products.
Such decisions are largely dependent upon the inherent complexity of a delivery infrastructure. Dimensions include:
■ Number and type of underlying technology (legacy to bleeding edge)
■ Inherent application complexity
■ Extensibility – physical or virtual
■ 3rd party inclusions (client side affiliates to server-side web services links)
■ Nature and distribution of end usage – devices, native applications, geography, etc.
It is worth considering the overall benefits of (for example) gaining say 75% of the theoretical maximal visibility and value from an easily configurable and accessible "light touch" APM; against driving for maximum value and precision. The latter approach necessarily requires higher overheads in a range of areas, from more intricate definition of output dashboards/reports to greater ongoing management and interpretative skills. Whatever the salesman says, none of the higher end APM tools are truly "plug and play" – all require tuning, both to provide good understanding and to minimize performance overhead.
As always, the ability of your chosen APM to work with your current (and anticipated) technology stack – application framework, legacy components (mainframe, VAX?), modern extensions (microservices containers), in addition to the ability to cope with your anticipated throughput volumes and demand patterns) are primary requirements.
Larry Haig is Senior Consultant at Intechnica.