In this section, we highlight select key requirements for NPM & APM platforms based on use cases and end-user demands. Although today’s NPM and APM platforms have progressed, they are continuing to address the problems of the past, including how to efficiently monitor an increasingly diverse and complex set of objects, cope with exponentially growing bandwidths, and the growing need to enable a broader set of management systems with accurate NPM and APM data.
Data Collection and Storage
An NPM/APM platform is only as valuable as the breadth of devices, technologies, applications, and objects that it can monitor. Many organization who have deployed NPM have done so solely to leverage the data collection and aggregation capabilities, and the resultant performance data that they maintain.
In the past, data collection was performed by polling devices directly (frequently via SNMP), or collecting data batch mode from network management platforms. This imposed a few key tradeoffs:
- Data Collection Plug-ins – NPM and APM users required a ‘Plug-In’ or ‘Collectors” for each type of metric of interest. Typically, the end-user must obtain the Plug-In/Collector from the tool vendor. Some vendors provide tools to semi-automate the process of adding new device support for their platforms. Others have rendered the process as a services engagement.
- Collection Rates vs. Data Collection Overhead – Ideally, NPM platforms could collect and present highly granular statistics in near-real time. However, the growing number of devices, managed objects, and bandwidths compel the need to tradeoff the frequency of data collection against the data collection overhead incurred. For polled systems, where a platform polls individual devices at regular intervals, the overhead affects the control processor, which is already burdened. For passive monitoring systems, the tradeoffs is the number of data collection points needed (and associated cost) to ensure the optimal
Another area that is particularly challenging is monitoring for cloud services, and virtualized network overlays, and SDN, which requires the ability to monitor tunnel and service endpoints, as encrypted payloads cannot be natively monitored. Some firms have begun to integrate their offerings directly with the Hyperscale Cloud operators, a trend that will no doubt expand as Public Cloud adoption continues to expand.