In this paper we propose a framework based on Deep Reinforcement Learning to proactively and autonomously take under control links loads in Segment Routing (SR) networks. The main idea is to monitor ...local link loads and, in case of anomalous situation, to execute local routing changes at milliseconds timescale. The solution proposed is based on a Multi Agent Reinforcement Learning (MARL) approach: a subset of nodes is equipped with a local agent, powered by a Deep Q-Network (DQN) algorithm, referred to as SRv6 rerouting for Local In-network Link Load Control (SR-LILLC). The main feature of SR-LILLC is to train the agents in a collaborative way, by defining a "shared" reward function, while working in an independent way during the operating phase. Moreover, the re-routing operation is performed in a transparent way for other network devices, without involving the centralized control plane, by exploiting the source routing feature of the SR. The performance evaluation conducted over real data sets shows that SR-LILLC is able to reduce the load on agents links without increasing the maximum link utilization of the network; moreover, the overall network performance are improved in terms of end-to-end delays.
In this paper we propose a framework based on Deep Reinforcement Learning to proactively and autonomously maintain the links load of a Segment Routing (SR) node under a pre-fixed threshold at ...milliseconds time scale. A local agent, powered by a Deep Q-Network (DQN) algorithm, referred to as intelligent Link Load Control (iLLC), selects are-routing operation to move traffic flows from overloaded links to alternative paths. The re-routing operation is performed in a transparent way for other network devices, without involving the centralized control plane, by exploiting the source routing feature of the SR. The performance evaluation conducted over real data sets shows that iLLC is able to distribute traffic load peaks o ver locallinks without degrading the overall network performance. Furthermore, iLLC outperforms a higher complexity heuristic based on capacity constraints checking, since it is able to select rerouting operations not impacting the global maximum link utilization.
In this paper we introduce QLR, a per-router control agent that aims at reducing the occupancy of the local buffers by performing re-routing operations. The Segment Routing architecture is exploited ...to manage the uncoordinated selection of re-routing performed by different nodes, thus avoiding the creation of routing loops, while the Extensible In-band Processing is used to allow the network nodes to have a detailed and updated view of the wide network status. Data and control plane programmability are considered to define a prototype implementation of QLR that allows for the execution of a preliminary performance evaluation and proof-of-concept. From the conducted experiments has emerged that QLR can effectively reduce the maximum queue occupancy and end-to-end delay up to 43% and 63%, respectively.
In recent years, a higher number of emerging network applications boosted the adoption of new managing technology, like Digital Twin (DT). Also in networking scope, DT starts to be largely considered ...to model network behaviours to efficiently manage their resources. DT can be exploited to find optimal solutions, even in context where disjoint objectives are envisaged. For instance, this is the case of network application handled by an SDN Controller that pursue possibly conflicting goals. Nowadays, SDN Controller has no feature to manage this kind of contentions. This paper proposes a novel module, called Digital Twin Manager, that aims at evaluating the effects of operational changes provided by concurrent network applications by means of a Digital Twin Network, i.e., an interconnection of several DTs. DT Manager updates the digital network status in real-time and evaluates the feedback information provided by the digital applications. The proof-of-concept shows the feasibility of the proposed solution.