Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Jeff J., Chung, Jae-Won, Ahn, Jisang, Liang, Yizhuo, Lu, Runyu, Jajoo, Akshay, Lee, Myungjin, Chowdhury, Mosharaf
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910132327153664
author Ma, Jeff J.
Chung, Jae-Won
Ahn, Jisang
Liang, Yizhuo
Lu, Runyu
Jajoo, Akshay
Lee, Myungjin
Chowdhury, Mosharaf
author_facet Ma, Jeff J.
Chung, Jae-Won
Ahn, Jisang
Liang, Yizhuo
Lu, Runyu
Jajoo, Akshay
Lee, Myungjin
Chowdhury, Mosharaf
contents Any-to-Any models are an emerging class of multimodal models that accept combinations of text and multimodal data as input and generate them as output, introducing heterogeneous computation paths and component scaling characteristics. There are existing mechanisms for deploying Any-to-Any models--or special cases of them--for inference serving, but they either require manual effort and expertise to tune, or do not generalize to generic Any-to-Any models. We present Cornfigurator, the first deployment planner for generic Any-to-Any model inference serving. The goal of Cornfigurator is to maximize the overall goodput of serving the model, defined as the throughput of requests meeting their latency targets. To do so, based on model and workload characteristics, Cornfigurator explores the full spectrum of deployment strategies, from colocation to disaggregation and mixing different strategies. Cornfigurator performs coarse-to-fine statistical evaluation to efficiently navigate the large space of candidate plans. Plans generated by Cornfigurator either match or deliver 1.12$\times$-6.32$\times$ higher goodput compared to existing systems and expert-tuned deployment plans.
format Preprint
id arxiv_https___arxiv_org_abs_2512_14098
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving
Ma, Jeff J.
Chung, Jae-Won
Ahn, Jisang
Liang, Yizhuo
Lu, Runyu
Jajoo, Akshay
Lee, Myungjin
Chowdhury, Mosharaf
Machine Learning
Distributed, Parallel, and Cluster Computing
Any-to-Any models are an emerging class of multimodal models that accept combinations of text and multimodal data as input and generate them as output, introducing heterogeneous computation paths and component scaling characteristics. There are existing mechanisms for deploying Any-to-Any models--or special cases of them--for inference serving, but they either require manual effort and expertise to tune, or do not generalize to generic Any-to-Any models. We present Cornfigurator, the first deployment planner for generic Any-to-Any model inference serving. The goal of Cornfigurator is to maximize the overall goodput of serving the model, defined as the throughput of requests meeting their latency targets. To do so, based on model and workload characteristics, Cornfigurator explores the full spectrum of deployment strategies, from colocation to disaggregation and mixing different strategies. Cornfigurator performs coarse-to-fine statistical evaluation to efficiently navigate the large space of candidate plans. Plans generated by Cornfigurator either match or deliver 1.12$\times$-6.32$\times$ higher goodput compared to existing systems and expert-tuned deployment plans.
title Cornfigurator: Automated Planning for Any-to-Any Multimodal Model Serving
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2512.14098