George Parisis

Luca Giacomoni, Basil Benny, George Parisis

Reinforcement Learning (RL) has gained significant momentum in the development of network protocols. However, RL-based protocols are still in their infancy, and substantial research is required to build deployable solutions. Developing a protocol based on RL is a complex and challenging process that involves several model design decisions and requires significant training and evaluation in real and simulated network topologies. Network simulators offer an efficient training environment for RL-based protocols, because they are deterministic and can run in parallel. In this paper, we introduce \textit{RayNet}, a scalable and adaptable simulation platform for the development of RL-based network protocols. RayNet integrates OMNeT++, a fully programmable network simulator, with Ray/RLlib, a scalable training platform for distributed RL. RayNet facilitates the methodical development of RL-based network protocols so that researchers can focus on the problem at hand and not on implementation details of the learning aspect of their research. We developed a simple RL-based congestion control approach as a proof of concept showcasing that RayNet can be a valuable platform for RL-based research in computer networks, enabling scalable training and evaluation. We compared RayNet with \textit{ns3-gym}, a platform with similar objectives to RayNet, and showed that RayNet performs better in terms of how fast agents can collect experience in RL environments.

Mihai Mazilu, Luca Giacomoni, George Parisis

Learning-based congestion control (CC), including Reinforcement-Learning, promises efficient CC in a fast-changing networking landscape, where evolving communication technologies, applications and traffic workloads pose severe challenges to human-derived, static CC algorithms. Learning-based CC is in its early days and substantial research is required to understand existing limitations, identify research challenges and, eventually, yield deployable solutions for real-world networks. In this paper, we extend our prior work and present a reproducible and systematic study of learning-based CC with the aim to highlight strengths and uncover fundamental limitations of the state-of-the-art. We directly contrast said approaches with widely deployed, human-derived CC algorithms, namely TCP Cubic and BBR (version 3). We identify challenges in evaluating learning-based CC, establish a methodology for studying said approaches and perform large-scale experimentation with learning-based CC approaches that are publicly available. We show that embedding fairness directly into reward functions is effective; however, the fairness properties do not generalise into unseen conditions. We then show that RL learning-based approaches existing approaches can acquire all available bandwidth while largely maintaining low latency. Finally, we highlight that existing the latest learning-based CC approaches under-perform when the available bandwidth and end-to-end latency dynamically change while remaining resistant to non-congestive loss. As with our initial study, our experimentation codebase and datasets are publicly available with the aim to galvanise the research community towards transparency and reproducibility, which have been recognised as crucial for researching and evaluating machine-generated policies.

Aiden Valentine, Ian Wakeman, George Parisis

The highly dynamic nature of Low-Earth Orbit (LEO) satellite networks introduces challenges that existing transport protocols fail to address, including non-congestive latency variation and loss, transient congestion hotspots, and frequent handovers that cause temporary disconnections and route changes with unknown congestion and delay characteristics. Our contention is that with this increase in complexity, there is insufficient information being returned from the network for existing congestion control algorithms to minimise latency while maintaining high throughput and minimising retransmissions. Our approach, OrbCC, leverages in-network support to collect per-hop congestion information and uses it to (1) minimise buffer occupancy and end-user latency, (2) maximise application throughput and network utilisation, and (3) rapidly respond to congestion hotspots. We evaluate OrbCC against state-of-the-art transport protocols using OMNeT++/INET-based LEO satellite simulations and targeted micro-benchmarks. The simulations capture RTT dynamics in a LEO constellation, while the micro-benchmarks isolate key characteristics such as non-congestive latency variation and loss, path changes, and congestion hotspots. Results show that OrbCC significantly improves goodput while simultaneously reducing latency and retransmissions compared to existing approaches.

Giles Winchester, George Parisis, Luc Berthouze

Microservices have transformed software architecture through the creation of modular and independent services. However, they introduce operational complexities in service integration and system management that makes swift and accurate anomaly detection and localisation challenging. Despite the complex, dynamic, and interconnected nature of microservice architectures, prior works that investigate metrics for anomaly detection rarely include explicit information about time-varying interdependencies. And whilst prior works on fault localisation typically do incorporate information about dependencies between microservices, they scale poorly to real world large-scale deployments due to their reliance on computationally expensive causal inference. To address these challenges we propose FC-ADL, an end-to-end scalable approach for detecting and localising anomalous changes from microservice metrics based on the neuroscientific concept of functional connectivity. We show that by efficiently characterising time-varying changes in dependencies between microservice metrics we can both detect anomalies and provide root cause candidates without incurring the significant overheads of causal and multivariate approaches. We demonstrate that our approach can achieve top detection and localisation performance across a wide degree of different fault scenarios when compared to state-of-the-art approaches. Furthermore, we illustrate the scalability of our approach by applying it to Alibaba's extremely large real-world microservice deployment.

Mihai Mazilu, Aiden Valentine, George Parisis

Low Earth Orbit (LEO) satellite networks introduce unique congestion control (CC) challenges due to frequent handovers, rapidly changing round-trip times (RTTs), and non-congestive loss. This paper presents the first comprehensive, emulation-driven evaluation of CC schemes in LEO networks, combining realistic orbital dynamics via the LeoEM framework with targeted Mininet micro-benchmarks. We evaluated representative CC algorithms from three classes, loss-based (Cubic, SaTCP), model-based (BBRv3), and learning-based (Vivace, Sage, Astraea), across diverse single-flow and multi-flow scenarios, including interactions with active queue management (AQM). Our findings reveal that: (1) handover-aware loss-based schemes can reclaim bandwidth but at the cost of increased latency; (2) BBRv3 sustains high throughput with modest delay penalties, yet reacts slowly to abrupt RTT changes; (3) RL-based schemes severely underperform under dynamic conditions, despite being notably resistant to non-congestive loss; (4) fairness degrades significantly with RTT asymmetry and multiple bottlenecks, especially in human-designed CC schemes; and (5) AQM at bottlenecks can restore fairness and boost efficiency. These results expose critical limitations in current CC schemes and provide insight for designing LEO-specific data transport protocols.

Md Ashfaq Salehin, George Parisis, Luc Berthouze

Temporal random walks, which sample causality-preserving paths, are widely used to analyze time-stamped interactions in domains such as microservices, finance, and online platforms. Generating such walks at scale is challenging because real-world graphs evolve as high-volume streams, making continuous ingestion, efficient memory usage, and strict temporal ordering essential for practical deployment. We present Tempest (TEMPoral nEtwork Streaming Traversals), a GPU-accelerated engine for streaming temporal random walks. Tempest combines a GPU-native dual-index organization over a shared edge store with a hierarchical cooperative scheduler that dispatches walks at thread, warp, or block granularity based on per-step node convergence, enabling efficient start-edge selection, hop-by-hop causality enforcement, and window-based eviction without synchronization. It further provides closed-form constant-time samplers for common temporal bias functions. Our evaluation demonstrates sustained real-time processing of billion-edge streams under sliding windows, outperforming prior systems in ingestion and walk generation throughput while preserving causal correctness.

Antoine Messager, George Parisis, Istvan Z Kiss et al.

We consider the problem of inferring the functional connectivity of a large-scale computer network from sparse time series of events emitted by its nodes. We do so under the following three domain-specific constraints: (a) non-stationarity of the functional connectivity due to unknown temporal changes in the network, (b) sparsity of the time-series of events that limits the effectiveness of classical correlation-based analysis, and (c) lack of an explicit model describing how events propagate through the network. Under the assumption that the probability of two nodes being functionally connected correlates with the mean delay between their respective events, we develop an inference method whose output is an undirected weighted network where the weight of an edge between two nodes denotes the probability of these nodes being functionally connected. Using a combination of windowing and convolution to calculate at each time window a score quantifying the likelihood of a pair of nodes emitting events in quick succession, we develop a model of time-varying connectivity whose parameters are determined by maximising the model's predictive power from one time window to the next. To assess the effectiveness of our inference method, we construct synthetic data for which ground truth is available and use these data to benchmark our approach against three state-of-the-art inference methods. We conclude by discussing its application to data from a real-world large-scale computer network.