TAO: Tool-Aware Joint Scheduling for LLM-Based Agentic Serving
Shuhuai Li, Shengdong Ye, Dongdong Ge, Yinyu Ye, Jianghao Lin · AIOR
Collected · snapshot shipped with the site build, not a live feed
Abstract · as written by the authors
Agentic services alternate between model inference on clustered GPUs and external tools on distributed CPUs. In this scenario, existing systems (1) overlook tool-return readiness in service evaluation, wasting limited resources, and (2) use heuristic strategies to decide admission and placement on GPUs, resulting in suboptimal scheduling decisions. To bridge this gap, we present TAO, a tool-aware joint scheduling framework for LLM-based agentic serving. Its value interface combines near-term tool readiness, node-specific cache credit with time-aware decay, and token footprint; a dynamic program jointly selects and places waiting sessions under capacity constraints. Across the 12 reported model-workload-concurrency settings, TAO reduces latency by up to 26.8%, reduces mean time to first token (TTFT) by up to 36.3%, and increases decode throughput by up to 29.5% relative to the best evaluated state-of-the-art baseline for each metric.
This is the original abstract from arXiv. It has not been summarised, translated or rewritten.
Terms in plain language
AutomaticMade by matching the words of the title and abstract against a fixed list of definitions. It does not read the paper and does not change the abstract above.
- LLM
- an AI system trained on very large amounts of text to read and write language.
- Agentic
- an AI system that plans and takes several steps, often by calling tools, towards a goal.
- token
- a small unit that a model reads and writes, often a part of a word.
- state-of-the-art
- the best published result on a task so far, measured on a given benchmark.