arXiv subjects

Eytan Modiano

Publications and source records attributed to Eytan Modiano.

At least 19 recordsLinked to original sources

arXiv2026

Load Balancing with Partial Queue Information - Threshold Optimality and Indexability

We consider the problem of load balancing in a system with one dispatcher and $N$ parallel servers. The dispatcher must select one server to dispatch new jobs at every time-step and each server buffers incoming jobs in a queue. However, the dispatcher does not know the servers' backlogs and must make dispatching decisions based on previous observations. The dispatcher's objective is to dispatch jobs to the shortest queue. This problem can be formulated as a restless multi-arm bandit (RMAB) problem where each arm's state is its corresponding belief vector. Our goal is to verify Whittle indexability for this problem and derive a low complexity Whittle index policy. Previous Whittle indexability results cannot be directly applied due to the multi-dimensional nature of the belief vector. To overcome this issue, we define the RMAB state as the tuple of the most recent backlog observation and the time since this observation. We consider two model variations, a standard finite queue model and a blocking queue model. We show that the single-arm decoupled problems of both these models have threshold optimal solutions under some assumptions. For the standard finite queue model, we prove indexability and derive the Whittle index policy in closed form. For the blocking queue model, we derive a sufficient condition for indexability under threshold optimality and use it to show indexability for some special cases.

cs.NI↗

arXiv2026

Drift Plus Optimistic Penalty: A Learning Framework for Stochastic Network Optimization with Improved Regret Bounds

We consider the problem of joint routing and scheduling in queueing networks, where the edge transmission costs are unknown. At each time-slot, the network controller receives noisy observations of transmission costs only for those edges it selects for transmission. The network controller's objective is to make routing and scheduling decisions so that the total expected cost is minimized. This problem exhibits an exploration-exploitation trade-off, however, previous bandit-style solutions cannot be directly applied to this problem due to the queueing dynamics. In order to ensure network stability, the network controller needs to optimize throughput and cost simultaneously. We show that the best achievable cost is lower bounded by the solution to a static optimization problem, and develop a network control policy using techniques from Lyapunov drift-plus-penalty optimization and multi-arm bandits. We show that the policy achieves a sub-linear regret of order $O(\sqrt{T}\log T)$, as compared to the best policy that has complete knowledge of arrivals and costs. Finally, we evaluate the proposed policy using simulations and show that its regret is indeed sub-linear.

cs.NI↗

arXiv2026

Optimal Oblivious Load-Balancing for Sparse Traffic in Large-Scale Satellite Networks

Oblivious load-balancing in networks involves routing traffic from sources to destinations using predetermined routes independent of the traffic, so that the maximum load on any link in the network is minimized. We investigate oblivious load-balancing schemes for a $N\times N$ torus network under sparse traffic where there are at most $k$ active source-destination pairs. We are motivated by the problem of load-balancing in large-scale LEO satellite networks, which can be modelled as a torus, where the traffic is known to be sparse and localized to certain hotspot areas. We formulate the problem as a linear program and show that no oblivious routing scheme can achieve a worst-case load lower than approximately $\frac{\sqrt{2k}}{4}$ when $1<k \leq N^2/2$ and $\frac{N}{4}$ when $N^2/2\leq k\leq N^2$. Moreover, we demonstrate that the celebrated Valiant Load Balancing scheme is suboptimal under sparse traffic and construct an optimal oblivious load-balancing scheme that achieves the lower bound. Further, we discover a $\sqrt{2}$ multiplicative gap between the worst-case load of a non-oblivious routing and the worst-case load of any oblivious routing. The results can also be extended to general $N\times M$ tori with unequal link capacities along the vertical and horizontal directions.

cs.NI↗

arXiv2026

Scheduling in Multi-Hop Wireless Networks With Deadlines

We analyze the problem of scheduling in wireless networks to meet end-to-end service guarantees, defined by instantaneous throughput and hard packet deadlines. Using a network slicing model to decouple the queueing dynamics between flows, we show that the network's ability to meet hard deadline guarantees under interference is largely influenced by the link scheduling policy. We characterize throughput- and deadline-optimal policies for a solitary flow operating in isolation, which provide bounds on feasibility in the general case with multiple flows. We prove that packet delays can grow arbitrarily large in the multi-flow setting under a worst-case stabilizing policy, showing that queue stability is not sufficient to guarantee tight deadlines. We derive conditions on end-to-end packet delays in terms of link inter-scheduling times, and show that it is possible to make hard guarantees under any interference model by solving a generalized version of the pinwheel scheduling problem. Finally, we introduce a decentralized polynomial-time algorithm which can meet tight end-to-end packet deadlines while achieving near-optimal throughput.

cs.NI↗

arXiv2026

Utility Maximization in Wireless Backhaul Networks with Service Guarantees

We consider the problem of maximizing utility in wireless backhaul networks, where utility is a function of satisfied service level agreements (SLAs), defined in terms of end-to-end packet delays and instantaneous throughput. We model backhaul networks as a tree topology and show that SLAs can be satisfied by constructing link schedules with bounded inter-scheduling times, an NP-complete problem known as pinwheel scheduling. For symmetric tree topologies, we show that simple round-robin schedules can be optimal under certain conditions. In the general case, we develop a mixed-integer program that optimizes over the set of admission decisions and pinwheel schedules. We develop a novel pinwheel scheduling algorithm, which significantly expands the set of schedules that can be found in polynomial time over the state of the art. Using conditions from this algorithm, we develop a scalable, distributed approach to solve the utility-maximization problem, with complexity that is linear in the depth of the tree.

cs.NI↗

arXiv2025

Delay Optimization in a Simple Offloading System: Extended Version

We consider a computation offloading system where jobs are processed sequentially at a local server followed by a higher-capacity cloud server. The system offers two service modes, differing in how the processing is split between the servers. Our goal is to design an optimal policy for assigning jobs to service modes and partitioning server resources in order to minimize delay. We begin by characterizing the system's stability region and establishing design principles for service modes that maximize throughput. For any given job assignment strategy, we derive the optimal resource partitioning and present a closed-form expression for the resulting delay. Moreover, we establish that the delay-optimal assignment policy exhibits a distinct breakaway structure: at low system loads, it is optimal to route all jobs through a single service mode, whereas beyond a critical load threshold, jobs must be assigned across both modes. We conclude by validating these theoretical insights through numerical evaluation.

eess.SY↗

arXiv2025

Optimal Service Mode Assignment in a Simple Computation Offloading System: Extended Version

We consider a simple computation offloading model where jobs can either be fully processed in the cloud or be partially processed at a local server before being sent to the cloud to complete processing. Our goal is to design a policy for assigning jobs to service modes, i.e., full offloading or partial offloading, based on the state of the system, in order to minimize delay in the system. We show that when the cloud server is idle, the optimal policy is to assign the next job in the system queue to the cloud for processing. However, when the cloud server is busy, we show that, under mild assumptions, the optimal policy is of a threshold type, that sends the next job in the system queue to the local server if the queue exceeds a certain threshold. Finally, we demonstrate this policy structure through simulations.

eess.SY↗

arXiv2025

A QoS Framework for Service Provision in Multi-Infrastructure-Sharing Networks

We propose a framework for resource provisioning with QoS guarantees in shared infrastructure networks. Our novel framework provides tunable probabilistic service guarantees for throughput and delay. Key to our approach is a Modified Dirft-plus-Penalty (MDP) policy that ensures long-term stability while capturing short-term probabilistic service guarantees using linearized upper-confidence bounds. We characterize the feasible region of service guarantees and show that our MDP procedure achieves mean rate stability and an optimality gap that vanishes with the frame size over which service guarantees are provided. Finally, empirical simulations validate our theory and demonstrate the favorable performance of our algorithm in handling QoS in multi-infrastructure networks.

cs.NI↗

arXiv2025

Minimum-hop Constellation Design for Low Earth Orbit Satellite Networks

We consider a Low Earth Orbit (LEO) satellite network with each satellite capable of establishing inter-satellite link (ISL) connections for satellite-to-satellite communication. Since ISLs can be reoriented to change the topology, we optimize the topology to minimize the average shortest path length (ASPL). We characterize the optimal ASPL ISL topology in two families of topologies, 1) vertex-symmetric in which the ISL connections at a satellite node represent a motif that is repeated at all other satellite nodes, and 2) general regular topologies in which no such repeating pattern need exist. We establish ASPL lower bounds for both scenarios and show constructions for which they are achievable assuming each satellite makes 3 or 4 ISL connections. For the symmetric case, we show that the mesh grid is suboptimal in both ASPL and diameter. Additionally, we show there are constructions that maintain intra-orbital ISL connections while still achieving near-optimal ASPL performance. For the general case we show it is possible to construct networks with ASPL close to the general lower bound when the network is sufficiently dense. Simulation results show that for both scenarios, one can find topologies that are very close to the lower bounds as the network size scales.

cs.NI↗

arXiv2025

Centralized Versus Distributed Routing for Large-Scale Satellite Networks

An important choice in the design of satellite networks is whether the routing decisions are made in a distributed manner onboard the satellite, or centrally on a ground-based controller. We study the tradeoff between centralized and distributed routing in large-scale satellite networks. In particular, we consider a centralized routing scheme that has access to global but delayed network state information and a distributed routing scheme that has access to local but real-time network state information. For both routing schemes, we analyze the throughput and delay performance of shortest-path algorithms in networks with and without buffers onboard the satellites. We show that distributed routing outperforms centralized routing when the rate of changes in the network link state is comparable to the inherent propagation and transmission delays. In particular, we show that in highly dynamic networks without buffers, the distributed scheme achieves higher throughput than a centralized scheme. In networks with buffers, the distributed scheme achieves lower delays with the same throughput.

cs.NI↗

arXiv2025

A Novel Switch-Type Policy Network for Resource Allocation Problems: Technical Report

Deep Reinforcement Learning (DRL) has become a powerful tool for developing control policies in queueing networks, but the common use of Multi-layer Perceptron (MLP) neural networks in these applications has significant drawbacks. MLP architectures, while versatile, often suffer from poor sample efficiency and a tendency to overfit training environments, leading to suboptimal performance on new, unseen networks. In response to these issues, we introduce a switch-type neural network (STN) architecture designed to improve the efficiency and generalization of DRL policies in queueing networks. The STN leverages structural patterns from traditional non-learning policies, ensuring consistent action choices across similar states. This design not only streamlines the learning process but also fosters better generalization by reducing the tendency to overfit. Our works presents three key contributions: first, the development of the STN as a more effective alternative to MLPs; second, empirical evidence showing that STNs achieve superior sample efficiency in various training scenarios; and third, experimental results demonstrating that STNs match MLP performance in familiar environments and significantly outperform them in new settings. By embedding domain-specific knowledge, the STN enhances the Proximal Policy Optimization (PPO) algorithm's effectiveness without compromising performance, suggesting its suitability for a wide range of queueing network control problems.

cs.LG↗

arXiv2024

Queueing Delay Minimization in Overloaded Networks

We develop link rate control policies to minimize the queueing delay of packets in overloaded networks. We show that increasing link rates does not guarantee delay reduction during overload. We consider a fluid queueing model that facilitates explicit characterization of the queueing delay of packets, and establish explicit conditions on link rates that can minimize the average and maximum queueing delay in both single-hop and multi-stage (switching) networks. These min-delay conditions require maintaining an identical ratio between the ingress and egress rates of different nodes at the same layer of the network. We term the policies that follow these conditions rate-proportional policies. We further generalize the rate-proportional policies to queue-proportional policies, which minimize the queueing delay asymptotically based on the time-varying queue length while remaining agnostic of packet arrival rates. We validate that the proposed policies lead to minimum queueing delay under various network topologies and settings, compared with benchmarks including the backpressure policy that maximizes network throughput and the max-link-rate policy that fully utilizes bandwidth. We further remark that the explicit min-delay policy design in multi-stage networks facilitates co-optimization with other metrics, such as minimizing total bandwidth, balancing link utilization and node buffer usage. This demonstrates the wider utility of our main results in data center network optimization in practice.

cs.NI↗

arXiv2024

Fundamental Limits of Routing Attack on Network Overload

We quantify the threat of network adversaries to inducing \emph{network overload} through \emph{routing attacks}, where a subset of network nodes are hijacked by an adversary. We develop routing attacks on the hijacked nodes for two objectives related to overload: \emph{no-loss throughput minimization} and \emph{loss maximization}. The first objective attempts to identify a routing attack that minimizes the network's throughput that is guaranteed to survive. We develop a polynomial-time algorithm that can output the optimal routing attack in multi-hop networks with global information on the network's topology, and an algorithm with an approximation ratio of $2$ under partial information. The second objective attempts to maximize the throughput loss. We demonstrate that this problem is NP-hard, and develop two approximation algorithms with multiplicative and additive guarantees respectively in single-hop networks. We further investigate the adversary's optimal selection of nodes to hijack that can maximize network overload. We propose a heuristic polynomial-time algorithm to solve this NP-hard problem, and prove its optimality in special cases. We validate the near-optimal performance of the proposed algorithms over a wide range of network settings. Our results demonstrate that the proposed algorithms can accurately quantify the risk of overload given an arbitrary set of hijacked nodes and identify the critical nodes that should be protected against routing attacks.

cs.NI↗

arXiv2024

An Ordinary Differential Equation Framework for Stability Analysis of Networks with Finite Buffers

We consider the problem of network stability in finite-buffer systems. We observe that finite buffer may affect stability even in simplest network structure, and we propose an ordinary differential equation (ODE) model to capture the queuing dynamics and analyze the stability in buffered communication networks with general topology. For single-commodity systems, we propose a sufficient condition, which follows the fundamental idea of backpressure, for local transmission policies to stabilize the networks based on ODE stability theory. We further extend the condition to multi-commodity systems, with an additional restriction on the coupling level between different commodities, which can model networks with per-commodity buffers and shared buffers. The framework characterizes a set of policies that can stabilize buffered networks, and is useful for analyzing the effect of finite buffers on network stability.

cs.NI↗

arXiv2024

Optimal Slicing and Scheduling with Service Guarantees in Multi-Hop Wireless Networks

We analyze the problem of scheduling in wireless networks to meet end-to-end service guarantees. Using network slicing to decouple the queueing dynamics between flows, we show that the network's ability to meet hard throughput and deadline requirements is largely influenced by the scheduling policy. We characterize the feasible throughput/deadline region for a flow under a fixed route and set of slices, and find throughput- and deadline-optimal policies for a solitary flow. We formulate the feasibility problem for multiple flows in a general topology, and show its equivalence to finding a bounded-cost cycle on an exponentially large graph, which is unsolvable in polynomial time by the best-known algorithm. Using a novel concept called delay deficit, we develop a sufficient condition for meeting deadlines as a function of inter-scheduling times, and show that regular schedules are optimal for satisfying this condition. Motivated by this, we design a polynomial-time algorithm that returns an (almost) regular schedule, optimized to meet service guarantees for all flows.

cs.NI↗

arXiv2024

Achieving AoI Fairness in Spatially Distributed Wireless Networks: From Theory to Implementation

We design and implement two variants of a practical random access protocol called WiFair, based on IEEE 802.11 and designed to mitigate spatial unfairness in Age of Information (AoI). Drawing on previous theoretical work, we modify the mechanics of 802.11 to fairly minimize AoI in a wireless network consisting of several update nodes and a single base station. We implement this protocol on a testbed of software defined radios (SDRs) and measure its performance under a variety of settings compared to standard 802.11. We observe a 32% reduction in network average AoI and an 89% reduction in peak AoI in a last come first served (LCFS) single-packet queue setting, as well as a 76% reduction in network average AoI and an 82% reduction in peak AoI in a first come first served (FCFS) queue setting when the network is congested. We further show that when the network is uncongested, WiFair achieves the same performance as 802.11, and we demonstrate its robustness to more bursty traffic by streaming live video.

cs.NI↗

arXiv2024

Power Failure Cascade Prediction using Graph Neural Networks

We consider the problem of predicting power failure cascades due to branch failures. We propose a flow-free model based on graph neural networks that predicts grid states at every generation of a cascade process given an initial contingency and power injection values. We train the proposed model using a cascade sequence data pool generated from simulations. We then evaluate our model at various levels of granularity. We present several error metrics that gauge the model's ability to predict the failure size, the final grid state, and the failure time steps of each branch within the cascade. We benchmark the graph neural network model against influence models. We show that, in addition to being generic over randomly scaled power injection values, the graph neural network model outperforms multiple influence models that are built specifically for their corresponding loading profiles. Finally, we show that the proposed model reduces the computational time by almost two orders of magnitude.

eess.SY↗

arXiv2024

Intervention-Assisted Policy Gradient Methods for Online Stochastic Queuing Network Optimization: Technical Report

Deep Reinforcement Learning (DRL) offers a powerful approach to training neural network control policies for stochastic queuing networks (SQN). However, traditional DRL methods rely on offline simulations or static datasets, limiting their real-world application in SQN control. This work proposes Online Deep Reinforcement Learning-based Controls (ODRLC) as an alternative, where an intelligent agent interacts directly with a real environment and learns an optimal control policy from these online interactions. SQNs present a challenge for ODRLC due to the unbounded nature of the queues within the network resulting in an unbounded state-space. An unbounded state-space is particularly challenging for neural network policies as neural networks are notoriously poor at extrapolating to unseen states. To address this challenge, we propose an intervention-assisted framework that leverages strategic interventions from known stable policies to ensure the queue sizes remain bounded. This framework combines the learning power of neural networks with the guaranteed stability of classical control policies for SQNs. We introduce a method to design these intervention-assisted policies to ensure strong stability of the network. Furthermore, we extend foundational DRL theorems for intervention-assisted policies and develop two practical algorithms specifically for ODRLC of SQNs. Finally, we demonstrate through experiments that our proposed algorithms outperform both classical control approaches and prior ODRLC algorithms.

cs.AI↗