@inproceedings{yao-etal-2024-scalellm,
title = "{S}cale{LLM}: A Resource-Frugal {LLM} Serving Framework by Optimizing End-to-End Efficiency",
author = "Yao, Yuhang and
Jin, Han and
Shah, Alay Dilipbhai and
Han, Shanshan and
Hu, Zijian and
Stripelis, Dimitris and
Ran, Yide and
Xu, Zhaozhuo and
Avestimehr, Salman and
He, Chaoyang",
editor = "Dernoncourt, Franck and
Preo{\c{t}}iuc-Pietro, Daniel and
Shimorina, Anastasia",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = nov,
year = "2024",
address = "Miami, Florida, US",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-industry.22",
doi = "10.18653/v1/2024.emnlp-industry.22",
pages = "279--289",
abstract = "Large language models (LLMs) have surged in popularity and are extensively used in commercial applications, where the efficiency of model serving is crucial for the user experience. Most current research focuses on optimizing individual sub-procedures, e.g. local inference and communication, however, there is no comprehensive framework that provides a holistic system view for optimizing LLM serving in an end-to-end manner. In this work, we conduct a detailed analysis to identify major bottlenecks that impact end-to-end latency in LLM serving systems. Our analysis reveals that a comprehensive LLM serving endpoint must address a series of efficiency bottlenecks that extend beyond LLM inference. We then propose ScaleLLM, an optimized system for resource-efficient LLM serving. Our extensive experiments reveal that reveal that with 64 concurrent requests on Mixtral 8x7B, ScaleLLM achieves a 4.3{\mbox{$\times$}} speed up over vLLM and outperforms state-of-the-arts with 1.5{\mbox{$\times$}} higher throughput.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yao-etal-2024-scalellm">
<titleInfo>
<title>ScaleLLM: A Resource-Frugal LLM Serving Framework by Optimizing End-to-End Efficiency</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuhang</namePart>
<namePart type="family">Yao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Han</namePart>
<namePart type="family">Jin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alay</namePart>
<namePart type="given">Dilipbhai</namePart>
<namePart type="family">Shah</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shanshan</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zijian</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dimitris</namePart>
<namePart type="family">Stripelis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yide</namePart>
<namePart type="family">Ran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhaozhuo</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Salman</namePart>
<namePart type="family">Avestimehr</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chaoyang</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Franck</namePart>
<namePart type="family">Dernoncourt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniel</namePart>
<namePart type="family">Preoţiuc-Pietro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anastasia</namePart>
<namePart type="family">Shimorina</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, US</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large language models (LLMs) have surged in popularity and are extensively used in commercial applications, where the efficiency of model serving is crucial for the user experience. Most current research focuses on optimizing individual sub-procedures, e.g. local inference and communication, however, there is no comprehensive framework that provides a holistic system view for optimizing LLM serving in an end-to-end manner. In this work, we conduct a detailed analysis to identify major bottlenecks that impact end-to-end latency in LLM serving systems. Our analysis reveals that a comprehensive LLM serving endpoint must address a series of efficiency bottlenecks that extend beyond LLM inference. We then propose ScaleLLM, an optimized system for resource-efficient LLM serving. Our extensive experiments reveal that reveal that with 64 concurrent requests on Mixtral 8x7B, ScaleLLM achieves a 4.3\times speed up over vLLM and outperforms state-of-the-arts with 1.5\times higher throughput.</abstract>
<identifier type="citekey">yao-etal-2024-scalellm</identifier>
<identifier type="doi">10.18653/v1/2024.emnlp-industry.22</identifier>
<location>
<url>https://aclanthology.org/2024.emnlp-industry.22</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>279</start>
<end>289</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ScaleLLM: A Resource-Frugal LLM Serving Framework by Optimizing End-to-End Efficiency
%A Yao, Yuhang
%A Jin, Han
%A Shah, Alay Dilipbhai
%A Han, Shanshan
%A Hu, Zijian
%A Stripelis, Dimitris
%A Ran, Yide
%A Xu, Zhaozhuo
%A Avestimehr, Salman
%A He, Chaoyang
%Y Dernoncourt, Franck
%Y Preoţiuc-Pietro, Daniel
%Y Shimorina, Anastasia
%S Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, US
%F yao-etal-2024-scalellm
%X Large language models (LLMs) have surged in popularity and are extensively used in commercial applications, where the efficiency of model serving is crucial for the user experience. Most current research focuses on optimizing individual sub-procedures, e.g. local inference and communication, however, there is no comprehensive framework that provides a holistic system view for optimizing LLM serving in an end-to-end manner. In this work, we conduct a detailed analysis to identify major bottlenecks that impact end-to-end latency in LLM serving systems. Our analysis reveals that a comprehensive LLM serving endpoint must address a series of efficiency bottlenecks that extend beyond LLM inference. We then propose ScaleLLM, an optimized system for resource-efficient LLM serving. Our extensive experiments reveal that reveal that with 64 concurrent requests on Mixtral 8x7B, ScaleLLM achieves a 4.3\times speed up over vLLM and outperforms state-of-the-arts with 1.5\times higher throughput.
%R 10.18653/v1/2024.emnlp-industry.22
%U https://aclanthology.org/2024.emnlp-industry.22
%U https://doi.org/10.18653/v1/2024.emnlp-industry.22
%P 279-289
Markdown (Informal)
[ScaleLLM: A Resource-Frugal LLM Serving Framework by Optimizing End-to-End Efficiency](https://aclanthology.org/2024.emnlp-industry.22) (Yao et al., EMNLP 2024)
ACL
- Yuhang Yao, Han Jin, Alay Dilipbhai Shah, Shanshan Han, Zijian Hu, Dimitris Stripelis, Yide Ran, Zhaozhuo Xu, Salman Avestimehr, and Chaoyang He. 2024. ScaleLLM: A Resource-Frugal LLM Serving Framework by Optimizing End-to-End Efficiency. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 279–289, Miami, Florida, US. Association for Computational Linguistics.