Proactive Intent Policy Activation: An ML-Assisted Resource Forecasting Approach

Due to the mega-scale of next-generation networks, dynamic network policy management and assessment is a continuous struggle for network operators. As a result, networks have evolved towards automated Softwarized intent-based policy platforms. IBN (Intent-based Networking) administers high-level interpretation of intents and translates them as policies to the lower-level infrastructures. As a Softwarized platform, it defines an SSOT (Single Source of Truth) where every high-level intention has well-defined policy action at every stage of orchestration from top to bottom. However, due to the dynamic nature of the network, the static quantitative measures may result in excessive reservation or under-allocation of network resources. As network service orchestration is performed for the future, deciding on service parameters per resource availability is crucial. Hence, this work utilizes machine-learning-based resource forecasting to determine quantitative measures for intent activation proactively. This work empowers IBN policy activation through LSTM (Long Short-Term Memory) based compute resource forecasting and RouteNet-driven transport network path link utilization forecasting. The goal of IBN is to ensure service level agreement, as there are two primary service activation rules, GBR and Non-GBR (Guaranteed Bit Rate); hence using forecasting, the proposed model ensures the service agreement accomplishment proactively.