Observability is a hot topic in network operations these days. Actually, it has always been a hot topic but under different names such as monitoring, data collection, telemetry, etc. No matter what it’s called, the gathering of operational state data is crucial to all aspects of network operations. Performing this data collection in the most efficient manner is very important as well as having maximum visibility into useful data.
Historically, network operators have had to periodically poll the devices in their network in order to collect this information. This creates additional network traffic which may not be useful traffic. Additionally, it is not efficient in that the management systems doing the polling don’t know what is happening in a device and simply poll away on their selected cadence without regard to what may or may not be happening. Each time the device is polled, it has to spend computing resources and time gathering the data and responding to the request. This could be the exact same data that was gathered and sent in response to the exact same request during the previous polling period.
So, it is easy to see that a better way for collecting network operations data in a manner more efficient and useful than traditional polling is needed. This is where IETF’s work on YANG-Push comes into play. YANG-Push turns the traditional pull (i.e., polling) model into a push model. The device is told what data to send and when. Additionally, the “when” can be either periodic or on-change. This results in vastly improved efficiency in the gathering of network data.
ConfD’s first phase of YANG-Push support was introduced in ConfD 7.4. This support has grown in ConfD 7.5 and will continue to expand in future releases.
Our new application note “YANG-Push and ConfD” provides a basic introduction to this exciting new technology as well as an overview of YANG-Push support in ConfD as of the 7.5 release.