arXiv subjects

Shobha Rani

Publications and source records attributed to Shobha Rani.

2 recordsLinked to original sources

arXiv2026

Queueing Analysis and Cost Optimization in a Diagnostic--Treatment Hospital Queue with Heterogeneous Referred Patients

This study develops an analytical and decision framework for a pooled hospital service comprising first-time patients who require diagnosis followed by treatment and referred patients who require only their prescribed treatment mode. The system is represented as a Markovian phase-type queue. probability generating functions are derived for two-treatment cases, while a matrix-analytic method is developed for an arbitrary number $n$ of treatment modes. The key performance measures are derived and interpreted. A load-triggered control is designed to determine the minimum capacity increment required to achieve a prescribed utilization level. Numerical experiments are conducted to validate the analytical results. Furthermore, the total-cost problem admits a strictly convex reformulation in reciprocal service-time variables and has a unique global minimum. Particle Swarm Optimization (PSO), Simulated Annealing (SA), and the Sine Cosine Algorithm (SCA) are additionally used as independent heuristic solvers.

math.OC↗

arXiv2026

LLM Inference Under Bursty Workload Distribution: Modifying the WAIT Algorithm

Large Language Models (LLMs) such as ChatGPT and Claude are widely used for information retrieval and problem-solving. Recent work has focused on improving scheduling algorithms to boost throughput while maintaining low latency. However, these approaches often assume Poisson request arrivals with constant rates - an assumption that fails to reflect the inherently bursty and dynamic nature of real-world traffic. We propose a lightweight extension to the state-of-the-art WAIT algorithm [1], which adapts to time-varying arrival rates without prior traffic knowledge. The proposed algorithm performs online estimation of request intensity based on observed interarrival times. Using Markov Modulated Poisson Process (MMPP)-based synthetic workloads with diverse request types, we conduct a simulation-based evaluation demonstrating that the proposed method achieves higher throughput than Sarathi-Serve [2], ORCA [3], and vLLM [4] in the evaluated low arrival-rate shift scenarios while maintaining comparable latency.

cs.LG↗