This work assesses the reliability of a dispatch agent across 432 runs, and finds that descriptive metadata increases successful execution from 48% to 87% of 220 submitted jobs, eliminating architecture mismatch, and improving performance for five of ten measurable applications by up to 3.3x.
Abstract
Agents powered by AI/ML are becoming ingrained in orchestration. Dispatch of work is the task of receiving a request, transforming it for a workload manager, and successfully submitting it. Running scientific workflows across multi-cluster environments introduces substantial challenges of dynamic job transformation, dispatch, and submission to heterogeneous clusters. These tasks are well-suited to agents, which can receive textual instructions for work, prepare job specifications, and dispatch. In this work, we assess the reliability of a dispatch agent across 432 runs, testing all possible combinations of five feature dimensions across four prompt styles. The agent is highly reliable (97.9% success). We test a full orchestration to submit, queue, match, score, select, transform, and dispatch in a multi-cluster experiment. We find that descriptive metadata increases successful execution from 48% to 87% of 220 submitted jobs, eliminating architecture mismatch, and improving performance for five of ten measurable applications by up to 3.3x.
This paper presents a hierarchical, dynamic architecture and software to discover resources across diverse cloud, edge, and HPC systems and exemplifies the importance of careful coordination between agents, discovery tools, and infrastructure for agentic science.
This work presents MetaRoute-Bench, an open, inspectable framework for comparing meta-decision policies under a shared execution model, and releases task generation, policies, traces, tests, and analysis artifacts to support live-system validation.
Natan Vidra, Alina Kapanova, Arun Kanhai et al.· 0 citations
The emergence of modern agents powered by large language models has created a demand for executing long-horizon, autonomous workflows in various domains that require significant computational resources. While High Performance Computing clusters provide the ideal infrastructure for these computation-intensive workloads, traditional HPC job schedulers such as Slurm are not designed for dynamic, agentic workflows characterized by unpredictable task durations, external API calls, and fault tolerance requirements of modern agents. This work presents RASER, a user-space framework that enables seamless execution of agentic workflows on production HPC clusters by extending Slurm's internal primitives. RASER introduces agentic job arrays with work stealing via shared filesystem queues, user-space checkpointing through application-level state serialization combined with Slurm requeue, and Apptainer container-based isolation without requiring any image modifications. Evaluations demonstrate that RASER reduces makespan by nearly 39% compared to static partitioning while achieving near-full CPU utilization. RASER provides resilience against preemption and failures while maintaining minimal checkpoint/restore overhead. It requires no kernel privileges or external database infrastructure, making it an accessible solution for deploying agentic workflows on existing HPC infrastructure.
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, Hai-Ying Shen et al.· International Conference on...· 0 citations
This work organizes agentic workflows in a taxonomy and presents its first architectural characterization with a production study at Microsoft Azure and a controlled study of open-source frameworks, showing that agentic execution is fragmented and heterogeneous.
Ji-Rong Yang, Peizhe Liu, Chaojie Zhang et al.· 2 citations
Enterprise adoption of machine learning has fragmented operations into specialised disciplines—DataOps, MLOps, AIOps—creating silos that impede unified governance. We propose XOps, a five-layer reference architecture integrating PlatformOps, DataOps, MLOps and AIOps beneath an Agentic Orchestration layer with Policy-as-Code governance, together with a continuous-time Markov chain model quantifying the availability effect of agent-driven remediation. Two case studies evaluate the architecture under controlled, synthetic conditions. For a self-healing payment gateway, 250 live executions of the reasoning graph against a hosted language model and a live policy engine yield 85.6% plan-level action consistency and 99.6% fault classification accuracy, with 12.4% of plans rejected by the policy gate and escalated to a human and no policy-violating action authorised for execution; the pipeline’s 3.3-minute recovery time is a simulated latency budget rather than a cluster measurement. For a predictive-maintenance application on NASA C-MAPSS data, autonomous drift detection and retraining sustain
$$R^2 = 0.74$$
against 0.29 for an equivalent static model, measured on engines reserved entirely from retraining. An indicative cost analysis suggests an approximately 70% reduction in expected monthly operational cost. Within this scope the results support the feasibility of agent-driven operations rather than establishing production-scale performance.
Mete Köse, E. Küçüksille· Scientific Reports· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
