The cable access network has large amounts of monitoring data collected across various network equipment (CMTS, CMs, STBs, APs etc.). This data reflects the state of the network, status of devices and outside plant. Some examples are PNM data (RxMER, Channel coefficients), CM and CMTS MIBs (FEC stats, Service flow stats, packet drop counts), IPDR data, etc. Now, given this data set, how can the operator identify network plant issues, reliably, proactively and automatically?
Combining the analysis of data along, with network topology and device location, it is possible to create a general view of the plant condition and isolate problem sources. This paper tackles questions around, how can the operator identify network plant issues on a single modem, across a serving group, across fiber node, and correlate data across all devices in a reliable, proactive and automatic way. Machine-learning /data-analysis applications can crunch through layers of data to identify patterns and prioritize network issues automatically, allowing operators to reduce troubleshooting and problem resolution time, thereby reducing operational costs and enhance network reliability.