Conference
Open access
Sep 2026
SplitScaling: Adaptive Scaling for Disaggregated LLM Serving Against Traffic Bursts via DRL
This work proposes a Deep Reinforcement Learning-based auto-scaling framework tailored for the PD architecture that enables the agent to capture non-linear load dynamics, thereby achieving decoupled and precise scaling for prefill and decode pools.
Wei Xiao, Xue-Feng Huang, Wei-Jia Shi et al.
· Proceedings of the Thirty-Fi... · 0 citations