My son's online gaming habit (some might call it an addiction) illustrates a common issue in performance monitoring that carries over to the enterprise VoIP world. At home, I have no trouble responding to emails, but he may be sitting next to me cursing the performance of our home broadband because the ping times have gone from 30ms to 60ms! My ISP is monitoring the availability of the service, but doesn't see how bad the my son’s end-user experience can get.
That relationship is crucial.
People have a very low tolerance for poor quality voice services. Intermittent performance issues with other enterprise applications can be pretty tolerable, but any issue with a voice call or video conference is reported immediately. IT operations teams (who get those angry phone calls) recognize how critical their enterprise VoIP systems are, and invest in products to monitor them. However, most monitoring tools focus on the availability of the service and less on the end-user experience or service performance.
The most commonly deployed monitoring tools are supplied by vendors of the enterprise VoIP infrastructure. These monitoring tools report on the current state of the infrastructure: active calls, dropped calls, signalling errors, etc. But this is not monitoring end-user experience of the VoIP service. For that measurement, these infrastructure vendors use self-reported metrics such as mean opinion scores (MOS) from the handset. As with any self-reported metric, handset scores are suspect and likely to be, let's say, optimistic for call quality reporting.
IT operations teams also use synthetic calls to estimate call quality. This involves a product that can simulate a VoIP call between two endpoints and report on the quality of the call. Still, it is a simulated call without an actual end-user, and usually takes place at times without the network congestion that causes problems in the first. And again, this process only monitors the availability of the service—not the actual end-user experience. Also consider that VoIP technology for enterprises is capable of handling hundreds of simultaneous calls. One cannot assume that each of these calls will deliver exactly the same user experience as measured by the tool monitoring the simulated call.
A more comprehensive approach for monitoring enterprise VoIP and video conferences services for actual end-user experience involves analyzing the media packets flowing on the network itself. To achieve this, the solution must:
The real power of this approach is that every packet is captured, analyzed and tagged – so quickly finding and troubleshooting poor call quality becomes easier. Even using the packets to replay the actual call becomes possible, allowing the troubleshooting team to experience the call as the users did. Clicking from a call record to the underlying packet analysis enables understanding of to the root cause of poor call quality. When further drill-down to the actual packets is required, VoIP infrastructure teams can quickly assess the cause of the jitter and packet loss on the network.
For this approach to be truly useful, it needs to be scalable, simple, and cost effective to deploy. In the core of a large enterprise, there can be millions of packets per second requiring analysis that will automatically discover VoIP calls as they are set up, report performance metrics during the call and track the full lifecycle of the call to teardown. The solution must be architected for high-performance streaming analytics. The deployment requires cost effective options to ensure VoIP calls traversing remote segments of the network are also captured and analyzed.
Don't be fooled into thinking your organization is fully monitoring the performance of its enterprise VoIP and video services. From what we seen, this approach has only recently become scalable, simple, and cost effective to deploy to be useful for enterprises. As such, you may be missing the most important part of the picture: the end-user experience.
Check out our eBook: How VoIP teams use Corvil