Jul 2026· International Conference on Computer Communications and Networks· pp. 1-9· 0 citations· 23 references
Abstract
Multi-agent applications increasingly rely on shared large language model backends in the public cloud, where bursty workloads cause requests from different agents to contend for the same LLM instances, leading to long queues, memory imbalance, and severe tail-latency inflation. Existing approaches typically prioritize requests using coarse workflow positions or static execution heuristics, which fail to adapt to short-term overload dynamics. We present FlowGuard, a workflow-aware overload controller for multi-agent LLM serving. Its key insight is that under sustained overload, GPU cycles spent on requests whose execution service-level-objectives (SLOs) are already violated are wasted. FlowGuard continuously recomputes per-request slack and prioritizes requests with the greatest remaining time before their deadlines, thereby maximizing on-time completions. In addition, a resource-aware dispatcher jointly accounts for KV-cache memory pressure and in-flight queue depth to reduce preemption across shared instances. Evaluated on a deliberately over-subscribed two-GPU backend, where all policies exhibit high absolute miss rates (i.e., the percentage of workflows that miss their deadlines), FlowGuard reduces the miss rate by 14–28% points over workflow-oblivious and static-priority baselines under BurstGPT-driven load, and by 34–38% points under co-located mixed-agent workloads.
A Task-Oriented Prefix-Aware Scheduler that jointly decides which agent prefixes to keep in the cache and which requests to schedule for execution and scores candidate post-decision states by trading off the expected reduction in each task's longest remaining service path against the near-term benefit of downstream prefix reuse.
Hongqiu Ni, Han Tian, Chi Zhang et al.· 1 citation
The widespread adoption and strong generalizability of large language models (LLMs) lead to highly heterogeneous workloads that exhibit substantial variability in request lengths and latency requirements. This pronounced heterogeneity causes existing scheduling strategies to suffer from head-of-line blocking and inefficient request ordering, resulting in significant SLO violations and inflated end-to-end latency. In this paper, we present UNAS, a hierarchical scheduler for end-to-end SLO-oriented LLM serving. UNAS adopts a two-level design. At the cluster level, it classifies incoming requests into urgent and normal queues based on estimated service times derived from request features and runtime statistics, and dynamically updates their queue priorities as latency pressure evolves. At the instance level, UNAS performs fine-grained dispatch using a least-slack-first policy, prioritizing requests with the highest risk of SLO violations. Experimental results demonstrate that UNAS consistently outperforms representative baselines across diverse workloads, improving end-to-end SLO attainment by up to 11.1% and reducing mean end-to-end latency by up to $1.47 \times$.
Tian-Nan Fu, Jianxiong Liao, Xu Chen et al.· Fall Joint Computer Conferen...· 0 citations
As LLM agents increasingly rely on the Model Context Protocol (MCP) to invoke isolated external sandboxes, disaggregated sandbox deployment introduces a fundamental tension between resource utilization and interactive tail latency. Persistent long-lived sandbox reservations incur excessive memory overhead at scale, while lazy on-demand instantiation generates severe cold-start penalties that degrade response performance under multi-tenant, multi-turn agent workloads. To resolve this dilemma, we present SpecBox, a runtime built around speculative sandbox preallocation tailored for dynamic LLM agent execution pipelines. At its core, SpecBox implements keyword matching and streaming semantic embedding to enable intent-driven sandbox prewarming, which identifies pending tool execution demands mid-LLM token generation and fully overlaps sandbox bootstrapping with model inference. To extend prewarming windows across sequential agent steps, the framework leverages context-aware stochastic prefetching atop a sandbox dependency graph to probabilistically forecast future sandbox switches ahead of execution. We complement these speculative mechanisms with two orthogonal optimizations: a semantic result cache that prunes redundant repeated sandbox invocations, and a dedicated out-of-band shared-memory transport plane that bypasses conventional network serialization to deliver zero-copy artifact transfers. Evaluated on high-concurrency multi-turn agent traces, our prototype demonstrates that SpecBox cuts P99 end-to-end latency by up to $2.9\times$ relative to the on-demand sandbox baseline, while slashing peak memory consumption by $45.9\%$ compared to permanently reserved sandbox deployments.
Yihui Zhang, Tianyu Wo, Jinghao Wang et al.· arXiv.org· 0 citations
Public LLM services serve diverse multi-agent applications with varying workflow dependencies and performance requirements. Requests generated by these applications often exhibit commonality and interdependence, yet current systems largely ignore such application-level structure. As a result, at the LLM engine cluster level, assigning requests to engines with the shortest queue can cause inefficient KV-cache transfers across GPUs. Using three representative multi-agent applications, we show that current scheduling methods miss opportunities to (a) improve performance through KV-cache reuse and reduced data transfer, and (b) increase goodput via batch management informed by workflow dependencies. Motivated by these observations, we propose a Workflow-Aware Scheduling system for Multi-Agent LLM systems (WaSMa) that incorporates cluster-and engine-level scheduling to optimize LLM request execution across GPU resources. Experimental results show that WaSMa reduces the P95 tail latency by up to 48% and improves goodput by up to 107% compared to existing methods.
Uttam Rao, Ali Zafar Sadiq, Haiying Shen et al.· International Conference on...· 0 citations
Cascade, an LLM serving system that estimates and continuously updates this per-request latency budget from request characteristics, KV-cache state, and current system load, and uses a single per-request budget to jointly coordinate request scheduling and KV-cache management across the memory hierarchy.
Muhammad Adnan, R. Mahapatra, Prashant J. Nair et al.· 0 citations
Concurrent multi-agent workflows expose future dependencies and serving-state requirements while running on heterogeneous GPU pools with time-varying load, model residency, and resource availability. The logical workflow defines the required computation, whereas its physical scheduling units, model-lifecycle actions, resource ordering, and placement must be selected according to the observed pool state. We present a prediction-guided runtime that uses workflow forecasts to construct and optimize a physical execution graph. Predictor estimates device-specific activation latency, peak memory, and model-loading cost, then propagates these predictions through workflow dependencies to forecast activation readiness and future model demand. Constructor builds semantics-preserving fusion and model-lifecycle alternatives, while Scheduler jointly optimizes their selection, placement, and execution order based on the live pool state. Across a workload spanning three workflow scenarios on a heterogeneous GPU pool, our system reduces end-to-end makespan and overall p95 completion latency under burst arrivals by up to 36.8% and 25.9%, respectively, over state-of-the-art workflow schedulers. It also saves up to 24.63 GPU-s per completed session.
Jing-Hao Wang, Yi-Feng Zhang, Xiao Zhou et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
