A Novel Telemetry Compression Method for Enhancing Network Resilience

Building a network digital twin (NDT) of a virtual network (VN) and using it to analyze the potential what-if situations that might result from an automated decision allows a management system to avoid undesired VN situations. NDTs require as much monitoring information from the VN as possible, and this information must be as fine-grained as possible. Current telemetry systems cannot provide sufficient information without increasing the bandwidth available in the control plane for transmitting monitoring information. In this study, we investigated the structure of monitoring information to design lossy and lossless compression algorithms that can be used to maximize the amount of monitoring information encoded in each telemetry data unit. We evaluated our algorithms to demonstrate that our lossy algorithm can obtain, on average, a compression level higher than 99% with an accuracy higher than 80% and a compression level higher than 95% with an accuracy higher than 93%, whereas the lossless algorithm obtains compression levels of 33% and 20% when telemetry data are split into large and small chunks, respectively.