Brighten Godfrey

Gangmuk Lim, Wanyu Zhao, Brighten Godfrey et al.

Efficiently serving large language model (LLM) inference tasks is crucial both for user-perceived latency such as time-to-first-token (TTFT) and for GPU utilization. However, LLM request routing, that is, assigning each inference request to a GPU instance, is particularly challenging: execution is highly input-dependent; batching and KV-cache reuse create strong cross-request coupling; and latency responds nonlinearly to context length, model/engine settings, and heterogeneous accelerators. As a result, simple traditional load balancing algorithms, and even heuristics tailored for LLM inference, fail to achieve good performance. We present Lodestar, a novel learning-based request routing system for distributed GPU clusters. Lodestar continuously collects a snapshot of the cluster at per-request level, including real-time instance state, request characteristics, and observed performance, and trains an online reward predictor that it uses to route inference requests to the instance that will maximize given reward (e.g., minimizing TTFT). Lodestar is cloud-native and works seamlessly with existing serving stacks (vLLM). With continuous online adaptation to changing workloads and infrastructure conditions, Lodestar achieves 1.41x lower average TTFT and 1.47x lower P99 TTFT on average (up to 2.15x/1.86x on homogeneous and 4.38x/4.42x on heterogeneous clusters) compared to a state-of-the-art prefix cache and load-aware heuristic, and learns these efficient routing strategies within about 5 minutes, based on experiments in a public cloud GPU cluster.

Zhi Chen, Shehab Sarar Ahmed, Chenkai Wang et al.

Congestion controllers (CCs) are critical to network performance, and yet their robustness under adverse conditions remains insufficiently understood. While recent learning-based CCs have demonstrated strong performance in controlled environments, it is unclear how they compare to traditional CCs when controllers' input signals are corrupted or when environmental conditions become systematically challenging. In this paper, we introduce CCLab, an adversarial testing framework for systematically evaluating the robustness of both learning-based and non-learning-based CCs. CCLab includes a reinforcement learning (RL)-based adversarial agent that operates in a closed loop with the congestion control policy, generating bounded perturbations either on input signals (feature-level) or on external network conditions (environment-level), while preserving realism through explicit constraints. Using this framework, we compare learning-based CCs with non-learning-based CCs under both feature-level and environment-level adversarial conditions. While both types of CCs suffer from performance degradation under adversarial testing, we find that learning-based CCs, in general, are more robust than traditional human-designed algorithms. Finally, we show that our adversarial traces can be used to train more robust CCs that outperform existing learning-based CCs under both challenging and normal conditions.

Rahul Bothra, Alexander Krentsel, Saptarshi Mandal et al.

Traffic engineering (TE) has become a crucial tool for enforcing routing policy and maintaining operational efficiency in large networks. Existing TE solutions pick an objective function to optimize, aiming to balance (i) allocating traffic optimally with (ii) reacting quickly to demand changes and disruption events. However, as the scale of networks grows, the runtime of the existing optimal solution becomes infeasibly large. The alternative - approximate solvers - result in costly inefficiencies. We present GPU-Accelerated Traffic Engineering (GATE), which achieves the best of both worlds: enabling fast TE runtimes through a highly-parallelizable GPU-compatible decomposition, while iteratively converging to the provably optimal solution. GATE unlocks a unique set of desirable properties: it becomes increasingly parallelizable with network size, supports a wide spectrum of fairness objectives, and offers theoretically guaranteed convergence to the optimal solution and near-optimal convergence within a bounded time. We evaluate GATE on production traces from two large cloud WANs, and show that GATE achieves near-optimal solutions 5-10x faster than state-of-the-art.

Shehab Sarar Ahmed, William Sentosa, Yinjie Zhang et al.

Infrastructure protocols like Congestion Control (CC) seek to provide reliable performance across a wide range of Internet environments. Currently, protocol designers assess performance through hand-designed test cases or data sets captured from real environments. However, such approaches may inadvertently overlook critical facets of the algorithm's behavior when they encounter an unanticipated environment or workload. We seek to understand the unanticipated with \sys, a system that automatically generates adversarial network environments that cause a target protocol implementation to perform poorly. AdvNet employs machine learning-based optimization to generate environments, and incorporates a robust noise-handling mechanism to mitigate the variability inherent in real-world protocol performance. Although our approach is more general, this paper focuses specifically on transport protocols and their CC implementations. We showcase AdvNet's capability to create adversarial scenarios for 27 kernel-space implementations of both single-path and multi-path CC protocols, for several use cases with different performance goals. AdvNet identifies problematic network conditions that expose previously unnoticed Linux kernel bugs and uncovers hidden limitations in CC implementations, and provides insights about robustness. These results suggest that automated adversarial testing can be a valuable tool in protocol development, and that robustness is a useful new dimension for benchmarking CC protocols.

Pratyush Sahu, Rahul Bothra, Venkat Arun et al.

Designing the architecture of modern networked systems requires navigating a large, combinatorial space of hardware, systems, and configuration choices with complex cross-layer interactions. Architects must balance competing objectives such as performance, cost, and deployability while satisfying compatibility and resource constraints, often relying on scattered rules-of-thumb drawn from benchmarks, papers, documentation, and expert experience. This raises a natural question: can large language models (LLMs) reliably perform this kind of architectural reasoning? We find that they cannot. While LLMs produce plausible configurations, they frequently miss critical constraints, encode incorrect assumptions, and exhibit ``stickiness'' to familiar patterns. A natural workaround--iterative validation via simulation or experimentation--is often prohibitively expensive at scale and, in many cases, infeasible, particularly when comparing hardware-dependent alternatives. Motivated by this gap, we present Kepler, a lightweight reasoning framework for architecture design that combines structured, expert-driven specifications with SMT-based optimization. Kepler encodes architecturally significant properties--requirements, incompatibilities, and qualitative trade-offs--about systems, hardware, and workloads as constraints, and synthesizes feasible designs that optimize user-defined objectives. It operates at an abstract level, capturing ``rules-of-thumb'' rather than detailed system behavior, enabling tractable reasoning while preserving key interactions, and provides explanations for its decisions. Through experiments and case studies, we show that Kepler uncovers interactions missed by LLMs and supports systematic, explainable design exploration.

Sayed Hadi Hashemi, Sangeetha Abdu Jyothi, Brighten Godfrey et al.

The method of choice for parameter aggregation in Deep Neural Network (DNN) training, a network-intensive task, is shifting from the Parameter Server model to decentralized aggregation schemes (AllReduce) inspired by theoretical guarantees of better performance. However, current implementations of AllReduce overlook the interdependence of communication and computation, resulting in significant performance degradation. In this paper, we develop Caramel, a system that accelerates decentralized distributed deep learning through model-aware computation scheduling and communication optimizations for AllReduce. Caramel achieves this goal through (a) computation DAG scheduling that expands the feasible window of transfer for each parameter (transfer boundaries), and (b) network optimizations for smoothening of the load including adaptive batching and pipelining of parameter transfers. Caramel maintains the correctness of the dataflow model, is hardware-independent, and does not require any user-level or framework-level changes. We implement Caramel over TensorFlow and show that the iteration time of DNN training can be improved by up to 3.62x in a cloud environment.