systems administration

This work addresses the high cognitive burden on users and excessive carbon emissions associated with scientific computing due to the complexity of the SLURM job scheduler interface and its lack of energy-aware scheduling mechanisms. To mitigate these issues, the authors propose a modular Perl-based toolkit featuring a simplified command-line interface and a text-based user interface (TUI) that supports job monitoring, cancellation, and automatic generation of specialized submission scripts. A key innovation is the introduction of an “eco-mode” that enables automatic energy-efficient scheduling through off-peak workload shifting. This approach significantly lowers the usability barrier, enhances job management efficiency, and effectively reduces the carbon footprint of research computing workflows.

Containerized environments suffer from isolation uncertainty due to kernel sharing, undermining security guarantees across the container lifecycle. Method: This paper introduces the first structured threat modeling methodology tailored to the full container lifecycle—creation, runtime, and termination—extending the STRIDE framework to formally characterize verifiable attack surfaces at each stage. Leveraging systematic analysis of Linux namespaces and cgroups, we conduct empirical isolation testing. Contribution/Results: Our study identifies six distinct cross-container attack vectors inherent to shared-kernel architectures and quantitatively demonstrates that, under default configurations, PID and network namespace isolation fails in 37% of cases. The work delivers a reproducible, actionable theoretical framework and empirical benchmark for container security assessment—fundamentally diverging from conventional hypervisor-based virtualization security paradigms.

To address the challenges of difficult development and debugging, complex integration testing environments, and high pedagogical barriers in the GlideinWMS distributed system, this paper proposes the “Workspace Container” methodology—a unified, lightweight, containerized environment for development and education. Built upon a multi-container architecture—including Factory, Frontend, compute nodes, and batch systems—it integrates Docker and VS Code to enable one-click local deployment, offline debugging, and seamless IDE collaboration. The key contribution lies in abstracting development, testing, and teaching workflows into reusable, composable, standardized container units, thereby substantially reducing onboarding overhead for new users. Empirical validation across multiple workshops confirms that the full system runs efficiently on commodity laptops, accelerates development and debugging cycles, and significantly improves instructional interactivity and experimental reproducibility.

This paper investigates whether dynamic voltage and frequency scaling (DVFS) and server power-on/off scheduling exhibit synergistic effects for energy-efficient data center operation. Method: We formulate a joint control model based on a continuous-time Markov chain, integrating dual-speed operation (full-speed/low-speed) with non-negligible power-on/off delays, and optimize a weighted cost function combining response time and energy consumption. Contribution/Results: Theoretical analysis and numerical experiments demonstrate that, across the entire load spectrum, DVFS and power-state switching yield no significant synergy: each strategy dominates energy savings within distinct load regimes, and their combination reduces total cost by less than 3%. This work is the first to systematically reveal the inherent weak superposition—rather than multiplicative synergy—among mainstream server-level energy-saving mechanisms. It provides both theoretical justification and practical guidance for minimalist, principle-driven design of data center energy-efficiency policies.

This study addresses the lack of systematic guidance for enterprise software teams in choosing between monolithic and microservices architectures. The work proposes a decision-making framework that integrates technical and organizational factors, evaluating the trade-offs of each architecture across dimensions such as scalability, reliability, deployment efficiency, and organizational complexity. The assessment is grounded in system scale, business requirements, operational maturity, and long-term maintainability. Through architectural pattern analysis, a structured evaluation model, and multiple case studies, the authors develop a practical selection methodology tailored to real-world engineering contexts. This approach offers enterprises clear architectural evolution pathways and actionable guidelines aligned with their developmental stages, thereby significantly enhancing the rationality and sustainability of system design decisions.

This study addresses the inefficiency in serverless platforms caused by complex, non-conservative information flows among functions. It introduces Hodge decomposition—a novel application in this domain—to construct a service topology model that decomposes observed operational flows into locally correctable components and globally persistent harmonic modes. The work demonstrates that harmonic flows are intrinsic structural characteristics of the system rather than artifacts of misconfiguration, and leverages this insight to propose new optimization mechanisms such as the “dumping effect.” By constructing service flow spectra and performing harmonic analysis, the approach effectively identifies architectural-level performance bottlenecks, thereby validating its efficacy in uncovering structural inefficiencies and guiding targeted performance optimizations.

Traditional operating systems struggle to support goal-directed, dynamically tool-invoking agents with adaptive behaviors, exhibiting fundamental limitations in scheduling, state management, security, and observability. This work presents the first systematic design of an Agent Operating System (AOS) architecture, which introduces an agent control plane into conventional OS abstractions and rethinks core mechanisms—including scheduling, context management, capability registration, policy enforcement, and auditing. AOS clearly delineates responsibility boundaries and non-goals, establishing a multi-layered integration model spanning user-space runtimes to distributed control planes, thereby transcending the traditional OS assumption of deterministic program execution. The paper establishes novel system abstractions for agent-centric computing, proposes a security threat model and evaluation criteria, and makes significant advances in ensuring deterministic execution, auditability, and operational interpretability.

This study addresses operational inefficiencies, poor auditability, and upgrade challenges in legacy control systems of large-scale scientific facilities—such as CERN, Diamond Light Source, and Fermilab’s ACORN project. We propose a GitOps-based modernization framework that adopts Git as the single source of truth for declarative configurations and tightly integrates containerization, Infrastructure-as-Code (IaC), and cloud-native principles to establish an automated, traceable, and version-controlled control infrastructure. Notably, this work represents the first systematic integration of modern data pipelines and AI/ML capabilities into accelerator science control systems, enabling automated configuration deployment, closed-loop runtime telemetry, and intelligent anomaly detection. Empirical evaluation demonstrates significant improvements in system reliability, maintainability, and regulatory audit compliance. The approach provides a reusable technical paradigm and engineering framework for the digital transformation of big-science facilities.

This study addresses the underexplored phenomenon of software aging in GPU-accelerated large language model (LLM) services, particularly concerning memory leakage under heterogeneous software-hardware stacks and dynamic workloads. It pioneers the extension of software aging research into GPU-based LLM serving by conducting 216-hour stress tests across six co-deployment configurations, simultaneously monitoring multi-dimensional metrics from the host, GPU, and client perspectives. Employing time-series statistical analysis with autocorrelation correction and multiple hypothesis testing, the work systematically characterizes aging patterns. Significant memory aging is consistently observed across all configurations, revealing that leakage rates are highly sensitive to runtime environments and deployment settings. These findings confirm the prevalence and quantifiability of the issue and establish a reproducible framework bridging software aging and LLM service research.

This study addresses the lack of low-cost, easily deployable, and flexible solutions compatible with diverse Modbus sensors in current environmental monitoring systems. The authors propose an open-source data acquisition platform based on Raspberry Pi, featuring a custom-developed AtmosPyre library for Modbus communication. By integrating Ansible automation and YAML-based configuration, the system enables one-click deployment and modular driver extensions, significantly lowering the barrier to integrating new sensors. Utilizing an RS-485 interface, the platform stores data in CSV/JSON formats and achieves a hardware cost of only €54–63. It has been successfully deployed in a karst region for long-term, continuous monitoring of CO₂ and ²²²Rn since spring 2025, demonstrating stable operation. Both hardware and software are fully open-source, offering high cost-effectiveness and reproducibility.