about
Evolving Infra: Agent-Adaptive LLM Serving at Scale (arxiv.org)
1 point by matt_d 1 day ago | hide | past | pdf | discuss on HN

In plain words: A serving system for AI agents that lets GPU nodes switch between taking in input and writing replies, keeping the conversation memory agents will reuse rather than deleting it. Versus NVIDIA's serving framework, it cut the wait before a reply starts up to 14x.

Abstract

Agentic workloads are a major source of LLM inference demand. Existing prefill--decode (PD) disaggregated serving systems offer limited flexibility for runtime adaptation of compute allocation and KV cache management, and rely on coarse signals that only partly capture the diverse and dynamic agent workload. These limitations lead to load imbalance between prefill and decode nodes and eviction of KV caches that agents will reuse. We present Evolving Infra, an agent-adaptive serving framework that adapts compute allocation to agent workloads and KV cache management to agent context management at runtime. We decouple compute allocation from key--value (KV) cache management, allowing nodes to switch efficiently between prefill and decode. Building on this design, an online planner matches the PD split to workload composition, while an agent-aware distributed KV manager guides cache eviction and transfer with signals from agent context management. Across coding, math, and computer-use tasks with GLM-5.2 and Kimi K2.6 on up to 256 GB300 GPUs, Evolving Infra reduces mean end-to-end request latency and P95 TTFT by up to 4.1x and 14x over Dynamo, NVIDIA's production framework. We open-source the system implementation of Evolving Infra at https://github.com/mit-han-lab/EvolvingInfra.

Hao Kang, Weili Xu, Ziyang Li, Zheng Du, Hanchen Li, Tian Xia, Song Bian, Tushar Krishna, Huizi Mao, Ion Stoica, Song Han
arXiv:2611.12304 · cs.DC · submitted Nov 9, 2026 · updated Nov 12, 2026
abstract · pdf · html · We just modify some intro and add link to the previou paper

add comment on HN