Locality-aware Fair Scheduling in LLM Serving

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Cao, Shiyi, Wang, Yichuan, Mao, Ziming, Hsu, Pin-Lun, Yin, Liangsheng, Xia, Tian, Li, Dacheng, Liu, Shu, Zhang, Yineng, Zhou, Yang, Sheng, Ying, Gonzalez, Joseph, Stoica, Ion
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912202789748736
author Cao, Shiyi
Wang, Yichuan
Mao, Ziming
Hsu, Pin-Lun
Yin, Liangsheng
Xia, Tian
Li, Dacheng
Liu, Shu
Zhang, Yineng
Zhou, Yang
Sheng, Ying
Gonzalez, Joseph
Stoica, Ion
author_facet Cao, Shiyi
Wang, Yichuan
Mao, Ziming
Hsu, Pin-Lun
Yin, Liangsheng
Xia, Tian
Li, Dacheng
Liu, Shu
Zhang, Yineng
Zhou, Yang
Sheng, Ying
Gonzalez, Joseph
Stoica, Ion
contents Large language model (LLM) inference workload dominates a wide variety of modern AI applications, ranging from multi-turn conversation to document analysis. Balancing fairness and efficiency is critical for managing diverse client workloads with varying prefix patterns. Unfortunately, existing fair scheduling algorithms for LLM serving, such as Virtual Token Counter (VTC), fail to take prefix locality into consideration and thus suffer from poor performance. On the other hand, locality-aware scheduling algorithms in existing LLM serving frameworks tend to maximize the prefix cache hit rate without considering fair sharing among clients. This paper introduces the first locality-aware fair scheduling algorithm, Deficit Longest Prefix Match (DLPM), which can maintain a high degree of prefix locality with a fairness guarantee. We also introduce a novel algorithm, Double Deficit LPM (D$^2$LPM), extending DLPM for the distributed setup that can find a balance point among fairness, locality, and load-balancing. Our extensive evaluation demonstrates the superior performance of DLPM and D$^2$LPM in ensuring fairness while maintaining high throughput (up to 2.87$\times$ higher than VTC) and low per-client (up to 7.18$\times$ lower than state-of-the-art distributed LLM serving system) latency.
format Preprint
id arxiv_https___arxiv_org_abs_2501_14312
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Locality-aware Fair Scheduling in LLM Serving
Cao, Shiyi
Wang, Yichuan
Mao, Ziming
Hsu, Pin-Lun
Yin, Liangsheng
Xia, Tian
Li, Dacheng
Liu, Shu
Zhang, Yineng
Zhou, Yang
Sheng, Ying
Gonzalez, Joseph
Stoica, Ion
Distributed, Parallel, and Cluster Computing
Machine Learning
Large language model (LLM) inference workload dominates a wide variety of modern AI applications, ranging from multi-turn conversation to document analysis. Balancing fairness and efficiency is critical for managing diverse client workloads with varying prefix patterns. Unfortunately, existing fair scheduling algorithms for LLM serving, such as Virtual Token Counter (VTC), fail to take prefix locality into consideration and thus suffer from poor performance. On the other hand, locality-aware scheduling algorithms in existing LLM serving frameworks tend to maximize the prefix cache hit rate without considering fair sharing among clients. This paper introduces the first locality-aware fair scheduling algorithm, Deficit Longest Prefix Match (DLPM), which can maintain a high degree of prefix locality with a fairness guarantee. We also introduce a novel algorithm, Double Deficit LPM (D$^2$LPM), extending DLPM for the distributed setup that can find a balance point among fairness, locality, and load-balancing. Our extensive evaluation demonstrates the superior performance of DLPM and D$^2$LPM in ensuring fairness while maintaining high throughput (up to 2.87$\times$ higher than VTC) and low per-client (up to 7.18$\times$ lower than state-of-the-art distributed LLM serving system) latency.
title Locality-aware Fair Scheduling in LLM Serving
topic Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2501.14312