{"type":"ScholarlyArticle","title":"SplitScaling: Adaptive Scaling for Disaggregated LLM Serving Against Traffic Bursts via DRL","authors":[{"name":"Wei Xiao","slug":"wei-xiao","given":"Wei","family":"Xiao","affiliation":""},{"name":"Xuefeng Huang","slug":"xuefeng-huang","given":"Xuefeng","family":"Huang","affiliation":""},{"name":"Weijia Shi","slug":"weijia-shi","given":"Weijia","family":"Shi","affiliation":""},{"name":"Baokang Zhao","slug":"baokang-zhao","given":"Baokang","family":"Zhao","affiliation":""}],"abstract":"The disaggregated Prefill-Decode (PD) architecture has emerged as a prominent paradigm for efficient Large Language Model inference serving. However, resource management remains a critical challenge, particularly under the dual burstiness of real-world scenarios—characterized by volatile fluctuations in both request arrival rates and Prompt-to-Response ratios. Existing rule-based heuristics often fail to accurately identify system bottlenecks, leading to severe resource misallocation and Service Level Objective (SLO) violations. To address this, we propose a Deep Reinforcement Learning-based auto-scaling framework tailored for the PD architecture. By modeling the resource allocation problem as a Markov Decision Process, our framework enables the agent to capture non-linear load dynamics, thereby achieving decoupled and precise scaling for prefill and decode pools. Furthermore, to mitigate Head-of-Line blocking caused by scaling latency, we design an immediate rescheduling mechanism that migrates queued tasks to newly ready nodes in real-time. Experimental results driven by Azure bursty load traces demonstrate that our framework significantly reduces computational costs by 25.2% and 28.3% compared to the Static configuration and HeteroScale, respectively, while strictly adhering to SLOs.","doi":"10.24963/ijcai.2026/568","url":"https://ijcai.org/proceedings/2026/568","pdf":"https://ijcai.org/proceedings/2026/0568.pdf","markdown":"https://ijcai.org/proceedings/2026/568.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/568","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Machine Learning","track":"Main Track","number":568,"pages":{"first":5101,"last":5109},"keywords":["Planning and Scheduling: Learning in planning and scheduling","Machine Learning: Reinforcement learning"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}