Optimizing Distributed Deployment of Mixture-of-Experts Model Inference in Serverless Computing

With the advancement of serverless computing, running machine learning (ML) inference services over a serverless platform has been advocated, given its labor-free scalability and cost effectiveness. Mixture-of-Experts (MoE) models have been a dominant type of model architectures to enable large mode...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Annual Joint Conference of the IEEE Computer and Communications Societies S. 1 - 10
Hauptverfasser: Liu, Mengfan, Wang, Wei, Wu, Chuan
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 19.05.2025
Schlagworte:
Bayes methods
Computational modeling
Costs
Optimization
Pipelines
Prediction algorithms
Predictive models
Scalability
Serverless computing
Throughput
ISSN:2641-9874
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:With the advancement of serverless computing, running machine learning (ML) inference services over a serverless platform has been advocated, given its labor-free scalability and cost effectiveness. Mixture-of-Experts (MoE) models have been a dominant type of model architectures to enable large models nowadays, with parallel expert networks. Serving large MoE models on serverless computing is potentially beneficial, but has been underexplored due to substantial challenges in handling the skewed expert popularity and scatter-gather communication bottleneck in MoE model execution, for cost-efficient serverless MoE deployment and performance guarantee. We study optimized MoE model deployment and distributed inference serving on a serverless platform, that effectively predict expert selection, pipeline communication with model execution, and minimize the overall billed cost of serving MoE models. Especially, we propose a Bayesian optimization framework with multi-dimensional \epsilon\mathbf{-greedy} search to learn expert selection and optimal MoE deployment achieving optimal billed cost, including: 1) a Bayesian decision-making method for predicting expert popularity; 2) flexibly pipelined scatter-gather communication; and 3) an optimal model deployment algorithm for distributed MoE serving. Extensive experiments on AWS Lambda show that our designs reduce the billed cost of all MoE layers by at least 75.67% compared to CPU clusters while maintaining satisfactory inference throughput. As compared to LambdaML in serverless computing, our design achieves 43.41% lower cost with a throughput decrease of no more than 18.76%.
ISSN:2641-9874
DOI:10.1109/INFOCOM55648.2025.11044553