Adaptive GPU Resource Allocation for Multi-Agent Collaborative Reasoning in Serverless Environments

Guilin Zhang 1, Wulan Guo 1, Ziqi Tan 1 · 2025

Read Adaptive GPU Resource Allocation for Multi-Agent Collaborative Reasoning in Serverless Environments by Guilin Zhang 1, Wulan Guo 1, Ziqi Tan 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-22149.txt.

Description

Guilin Zhang, Wulan Guo, and Ziqi Tan present a system that hands out graphics processor time to teams of cooperating AI agents running on serverless platforms. Such multi-agent systems mix light coordinator agents with heavy specialist ones, and demand swings fast. Static schedules waste expensive hardware, while naive rotation leaves agents waiting.

Their allocator adjusts shares in real time from workload traits, agent priorities, and minimum resource guarantees, using an algorithm whose cost grows only linearly with the number of agents. In simulations of a four-agent workflow, the approach cut latency by about 85 percent compared with round-robin scheduling. It also held throughput close to static allocation while improving cost and hardware use.

Catalog record

CREATORGuilin Zhang 1, Wulan Guo 1, Ziqi Tan 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-22149

Browse the full library · About the Tower of Babel