From LLM Inference to Agentic Workloads: Characterization and Implications for Serving Systems

• URL: http://arxiv.org/abs/2608.15127v1
• Date: Sat, 15 Aug 2026 09:02:16 GMT
• Title: From LLM Inference to Agentic Workloads: Characterization and Implications for Serving Systems
• Abstract summary: Agentic applications are shifting AI serving from isolated model inference to long-running workloads in which LLMs coordinate tools, environments, and persistent state.<n>We present AgentSysBench, a benchmark suite and measurement toolkit with ten representative agentic applications and unified systems-level instrumentation.
• Score: 14.6340263288784
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Agentic applications are shifting AI serving from isolated model inference to long-running workloads in which LLMs coordinate tools, environments, and persistent state. However, the system behavior of these workloads---where latency, cost, and bottlenecks arise---remains poorly characterized, leaving serving systems to rely on assumptions built for conventional inference. We present AgentSysBench, a benchmark suite and measurement toolkit with ten representative agentic applications and unified systems-level instrumentation. Across controlled deployments and production traces, we identify six properties that distinguish agentic workloads from conventional LLM serving: (1) execution is heavyweight and stateful, with non-LLM components dominating latency in 5 of 10 applications and sandbox working-set memory peaking at 28 GB per session; (2) applications compose components with heterogeneous resource affinity---GPU-bound inference, memory-bound retrieval, CPU-bound sandboxes---whose task latencies diverge by up to 32x; (3) bottlenecks shift across requests, models, and deployments; (4) production sessions hold state idle for minutes to hours between active steps; (5) a control-plane tax---auxiliary LLM calls and context overhead from tool schemas and observations---crowds out productive compute and context; and (6) production traces from three applications reveal heavy cross-request redundancy in search queries and web fetches, exposing a large caching opportunity. Four design explorations demonstrate that these findings are actionable: task-aware serving reduces latency by 29--40%, communication-aware placement by up to 4.5x, state offloading reduces memory usage by 4.6x, and tool-result caching removes 35.2% of redundant search calls and saves 19.3% of aggregate search latency.

Related papers

• AgentR A Stateful and Recovery-Aware Software Architecture for LLM-based Auditable Workflows [14.042464830509864]
  AgentR represents research intent, generated queries, candidate paper assessments and gap analyses as durable workflow artifacts.<n>The design includes explicit processing state transitions, retries with exponential backoff, credit-aware pre-checks, ACID token-cost logging, and Type-2 slowly changing pricing records.<n>At the LLM stage, the system achieves 99.2% job completion, and mean latencies of 9.0 s, 18.9 s, and 25.4 s for intent decomposition, query generation, and paper scoring.
  arXiv  Detail & Related papers  (2026-08-15T14:51:43Z)
• Speculative Interaction Agents: Building Real-Time Agents with Asynchronous I/O and Speculative Tool Calling [64.40340291543971]
  We propose Speculative Interaction Agents to enable real-time interaction with small edge-scale models.<n>We also present a clock-based training methodology that adapts the model to handle streaming inputs and asynchronous responses.<n>This approach provides 1.6-2.2$times$ speedups with the Qwen2.5-3B-Instruct and Llama-3.2-3B-Instruct models across multiple tool calling benchmarks.
  arXiv  Detail & Related papers  (2026-05-13T11:20:52Z)
• LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues [80.29362825271768]
  LongMemEval-V2 is a benchmark for evaluating whether memory systems can help agents acquire the experience needed to become knowledgeable colleagues in customized environments.<n>LME-V2 contains 451 manually curated questions covering five core memory abilities for web agents.<n>We propose a suite of two memory methods: AgentRunbook-R, an efficient RAG-based memory with knowledge pools for raw state observations, events, and strategy notes, and AgentRunbook-C, which stores trajectories as files and invokes a coding agent to gather evidence in an augmented sandbox.
  arXiv  Detail & Related papers  (2026-05-12T17:59:34Z)
• SAGA: Workflow-Atomic Scheduling for AI Agent Inference on GPU Clusters [13.891522069967507]
  We propose a shift to program-level scheduling for compound AI workloads.<n>We present SAGA, a distributed scheduler that implements this abstraction through three mechanisms.
  arXiv  Detail & Related papers  (2026-05-01T09:05:28Z)
• MemSifter: Offloading LLM Memory Retrieval via Outcome-Driven Proxy Reasoning [78.46301394559903]
  Large Language Models (LLMs) are increasingly used for long-duration tasks.<n>Current methods face a trade-off between cost and accuracy.<n>MemSifter is a novel framework that offloads the memory retrieval process to a small-scale proxy model.
  arXiv  Detail & Related papers  (2026-03-03T02:57:38Z)
• AgentCgroup: Understanding and Controlling OS Resources of AI Agents [2.8139711959925244]
  AI agents are increasingly deployed in multi-tenant cloud environments, where they execute diverse tool calls within sandboxed containers.<n>We present a systematic characterization of OS-level resource dynamics in sandboxed AI coding agents.<n>Preliminary evaluation demonstrates improved multi-tenant isolation and reduced resource waste.
  arXiv  Detail & Related papers  (2026-02-10T02:37:42Z)
• Harli: SLO-Aware Co-location of LLM Inference and PEFT-based Finetuning on Model-as-a-Service Platforms [33.64527903547734]
  Harli improves the finetune throughput by 46.2% on average (up to over state-of-the-art serving systems)<n>Harli improves the finetune throughput by 46.2% on average (up to over state-of-the-art serving systems)
  arXiv  Detail & Related papers  (2025-11-13T05:58:52Z)
• Scalable Engine and the Performance of Different LLM Models in a SLURM based HPC architecture [3.746889836344766]
  This work elaborates on a High performance computing architecture based on Simple Linux Utility for Resource Management (SLURM)<n> Dynamic resource scheduling and seamless integration of containerized have been leveraged to manage CPU, GPU, and memory efficiently in multi-node clusters.<n>The obtained results pave ways for significantly more efficient, responsive, and fault-tolerant LLM inference on large-scale HPC infrastructures.
  arXiv  Detail & Related papers  (2025-08-25T09:11:27Z)
• MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents [84.62985963113245]
  We introduce MEM1, an end-to-end reinforcement learning framework that enables agents to operate with constant memory across long multi-turn tasks.<n>At each turn, MEM1 updates a compact shared internal state that jointly supports memory consolidation and reasoning.<n>We show that MEM1-7B improves performance by 3.5x while reducing memory usage by 3.7x compared to Qwen2.5-14B-Instruct on a 16-objective multi-hop QA task.
  arXiv  Detail & Related papers  (2025-06-18T19:44:46Z)
• ConServe: Fine-Grained GPU Harvesting for LLM Online and Offline Co-Serving [61.35068981176018]
  ConServe is a large language model (LLM) serving system that achieves high throughput and strong online latency guarantees.<n>We show that ConServe delivers an average of 2.2$times$ higher throughput and reduces online serving tail latency by 2.9$times$ on average compared to state-of-the-art systems.
  arXiv  Detail & Related papers  (2024-10-02T04:12:13Z)
• Fast Distributed Inference Serving for Large Language Models [12.703624317418237]
  We present FastServe, a distributed inference serving system for large language models (LLMs) FastServe exploits the autoregressive pattern of LLM inference to enable preemption at the granularity of each output token. We build a system prototype of FastServe and experimental results show that compared to the state-of-the-art solution vLLM, FastServe improves the throughput by up to 31.4x and 17.9x under the same average and tail latency requirements, respectively.
  arXiv  Detail & Related papers  (2023-05-10T06:17:50Z)
• NumS: Scalable Array Programming for the Cloud [82.827921577004]
  We present NumS, an array programming library which optimize NumPy-like expressions on task-based distributed systems. This is achieved through a novel scheduler called Load Simulated Hierarchical Scheduling (LSHS) We show that LSHS enhances performance on Ray by decreasing network load by a factor of 2x, requiring 4x less memory, and reducing execution time by 10x on the logistic regression problem.
  arXiv  Detail & Related papers  (2022-06-28T20:13:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.