Model-driven Telemetry for Cable Access Networks
Written by Sudhir Krishnan, Product Manager, Cable Access Business Unit
Telemetry literally means remote monitoring. Network monitoring has been utilized by operations teams for decades to keep an eye on the overall health of the network. The primary tool for network monitoring is Simple Network Management Protocol (SNMP). SNMP provides a standardized way to access remote network devices over IP network. Using Management Information Bases (MIB) published by IEFT the data models are standardized so device vendors can share data from their products with operators worldwide. SNMP managers poll network devices for data pertaining to specific MIBs based on the operations needs at any given time. While this has worked well over the many decades, SNMP has issues when the scale in terms of number of objects and network devices increases.
This led to the development of the NETCONF protocol for model-driven telemetry along with the YANG data modeling language. A paper, prepared by Brian Hedstrom of CableLabs and Colorado University, compares SNMP with NETCONF and states that as the number of managed objects goes beyond 500 the operational efficiency of NETCONF surpasses that of SNMP. NETCONF is able to handle 10,000 managed objects in a single transaction while SNMP’s best scenario is 2779 transactions for the same number of managed objects. There is an impact on the device CPU and memory usage due to these large number of transactions. CMTSs have a large number of managed objects as they manage RF and subscriber resources. It is not uncommon for a CMTS to have thousands of interface objects, for example. The impact of SNMP walks, therefore, is especially high for CMTS.
The complexity in cable networks has been increasing as new technology, such as DOCSIS 3.1 and Remote PHY, is introduced to satisfy increasing bandwidth needs of consumers and businesses. Cable operators also continue to add new services into their portfolios such as community wireless and small cell backhaul. In order to run a complex and diverse network (in terms of different kind of devices), cable operators need to collect data on the health of devices in near real-time basis. This data is used to analyze and predict future resource requirements as well as potential issues. Cable operators wish to adopt big-data philosophy for monitoring, where data from different parts of the network is collected into a Data Lake infrastructure. Different operations teams could run analytics applications directly on the collected data instead of accessing network devices individually.
Model-driven telemetry is the enabler of such a transformation. Devices push near real-time data on their operations through regular updates in a programmatically recognizable format (JSON, XML, GPB) which can then be directly stored in a database. Telemetry collectors are deployed in the headends and data centers to collect updates coming from network devices and store them in databases. There are a number of collectors already in the market, many of which are open source, Pipeline is one of them and was submitted to open-source by Cisco in March 2017.
Models for telemetry are defined in native YANG models and made public by network vendors. Standards organizations, IETF and OpenConfig, have defined models as well. Open source YANG compilers allow development of applications that can consume data sent by network devices in compliance with the YANG models.
The Cisco cBR-8, Cisco’s Evolved CCAP, will introduce model-driven telemetry in release 16.8.1 with support for some key DOCSIS and Openconfig/IETF models. Cisco’s Smartnode, the next generation RemotePHY node, will leverage telemetry to provide constant data on signal levels, spectrum and other key KPIs. With high visibility to operating conditions, the Cable operators can fine-tune their operations, leading to lower downtime and faster response to plant and network issues.
We already hear complaints from various operators that they are not able to scale their performance monitoring tools with SNMP and that, in fact, is becoming a hurdle to the deployment of new technology such as Remote PHY. Streaming model-driven telemetry aims to unlock the value that Remote PHY can provide to cable providers. It allows flexible data formats and use of open source tools to store and analyze the data. Model-driven telemetry allows an ecosystem of application developers to provide solutions to the specific operations needs of cable providers. Different service teams, within the organization, could develop and deploy applications specific to their business needs yet share data among themselves. API based data access allows for role-based authentication that can provide protection for sensitive data
More information on Model-Driven telemetry in these blogs!