Enterprise AI Analysis
Introduction to the Special Issue on Control of Computing Systems
Computing systems, from large-scale cloud data centers to resource-constrained embedded and edge platforms, increasingly operate under volatile workloads, heterogeneous hardware, and changing environments. Meeting performance goals while respecting hard constraints (e.g., power caps, thermal envelopes, latency budgets, availability targets, and reliability requirements) requires these systems to continuously adapt their configurations at run time. This has elevated feedback control loops from an implementation detail to a central architectural element in modern infrastructures, providing a principled way to deliver efficiency, safety, and robustness under uncertainty and disturbances. Whereas computing for control is a mature discipline, the reverse direction, i.e., control for computing systems, is comparatively recent. In this emerging area, control-theoretic concepts are used not merely as optimization heuristics, but as foundations for modeling and engineering self-adaptive and autonomic computing systems with analyzable behavior: stability, convergence, transient performance, robustness to modeling error, and constraint satisfaction. The aim of this special issue is to consolidate and advance this multidisciplinary perspective by gathering validated contributions that connect system modeling (including identification and estimation), controller synthesis and analysis, and implementation in realistic computing settings, with explicit attention to trustworthiness properties such as predictability, safety, and robust performance. The growing momentum of Control for Computing has been reflected in dedicated events within the control community (e.g., at CDC and CCTA) and in an increasing body of related work in software engineering and self-adaptation venues (e.g., ICSE, FSE, SEAMS, ACSOS). This special issue continues that trajectory by presenting contributions spanning data-center management, many-core processor control, proactive self-adaptation, decision-making under mixed observability, decentralized monitoring at the edge, and optimization methods for digital twins.
Published: 04 March 2026
Accepted: 25 February 2026
Revised: 25 February 2026
Received: 25 February 2026
Unlocking Peak Performance & Reliability
Applying advanced control theory principles to computing systems yields significant improvements in efficiency, stability, and adaptability.
0%
Efficiency Gains
0%
System Stability Index
0X
Adaptation Speed
0%
Resource Utilization
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
This special issue introduces the critical need for control systems in modern computing, addressing challenges from volatile workloads to hard constraints. It highlights the shift from control for computing to control of computing systems, emphasizing self-adaptive, autonomic systems with analyzable behaviors like stability, convergence, and robustness. The issue aims to advance interdisciplinary perspectives on system modeling, controller synthesis, and implementation, focusing on trustworthiness and robust performance.
The special issue focuses on interdisciplinary research at the intersection of computer science, software engineering, and control theory. Key areas include advanced dynamic modeling, control, and optimization for diverse computer systems, ranging from cloud to edge platforms. It delves into cybersecurity for control systems, self-protection, and resilience, as well as autonomous cloud and edge computing paradigms. Other critical themes encompass real-time and supervisory/discrete control for software, coordination of complex multiple feedback loops, and runtime estimation of software variability. These topics collectively aim to enhance predictability, safety, and robust performance in modern computing infrastructures.
This section provides a summary of the six highlighted contributions in this special issue, showcasing innovative approaches to control computing systems. These articles cover diverse applications, from optimizing ML performance and controlling many-core HPC processors to enabling proactive self-adaptation and managing NFR tradeoffs. They also explore decentralized monitoring in edge environments and advanced optimization for digital twins, demonstrating practical advancements and validated solutions.
Core Topics of the Special Issue
The special issue covers a range of interdisciplinary research at the intersection of computer science and engineering, software engineering, and control theory. Key topics include:
- Self-adaptive and self-organizing systems using control and machine learning
- Dynamic modeling, control, and optimization for computer systems
- Cybersecurity in control systems; self-protection and control for resilience
- Autonomous cloud and edge computing using control
- Real-time control systems
- Supervisory and discrete control for software systems
- Coordination of multiple feedback loops using multiple models
- Runtime estimation of software variability
OptimML: Joint Control of Inference Latency and Server Power Consumption for ML Performance Optimization
This paper proposes a MIMO control framework that jointly regulates inference latency and server power under power capping by treating the ML model size as an additional actuator. An adaptive scheme with online estimation and model switching is used to maintain control accuracy and stability under workload and hardware variations, and results are validated on a hardware testbed across common ML frameworks.
Modeling and Controlling Many-Core HPC Processors: an Alternative to PID and Moving Average Algorithms
The authors develop a detailed thermal-power model for modern heterogeneous many-core MPSoCs, capturing nonlinearities and coupling effects that impact control behavior. They introduce a fuzzy-inspired thermal capping strategy and an iterative voltage-selection method, showing improved thermal regulation and competitive performance via model-in-the-loop and hardware-in-the-loop co-simulations.
Context-Aware Proactive Self-Adaptation: A Two-Layer Model Predictive Control Approach
This work introduces a contextual goal model to represent how context influences requirements and priorities, and proposes a two-layer MPC mechanism for proactive adaptation. One layer performs context-aware predictive control, while a second layer adapts the controller objectives and constraints through requirements updates, yielding improved performance in experimental scenarios.
SPECTRA: A Markovian Framework for Managing NFR Tradeoffs in Systems with Mixed Observability
This paper addresses NFR trade-offs when some NFRs are fully observable while others are only partially observable. It proposes a multi-objective MOMDP-based decision framework and evaluates it on SDN-based remote data mirroring scenarios, demonstrating improved utility and planning efficiency relative to existing approaches.
DEMon: A Decentralized and Self-Adaptive Approach for Monitoring Volatile Edge Environments
The paper presents a decentralized monitoring system for edge environments based on stochastic gossip protocols, avoiding centralized bottlenecks and improving resilience. It includes a self-adaptive mechanism to tune monitoring parameters to balance monitoring QoS with resource consumption, and is validated through an implementation and experiments.
Bandit Neural Architecture Search for Digital Twin Optimisation: A Scientific Machine Learning Approach
Focusing on SciML surrogates for digital twins, the authors formulate NAS/HPO as a non-stochastic bandit problem to better match deterministic physics-based training regimes. They propose BanditNAS, analyze its theoretical properties, and report empirical results across representative SciML settings, highlighting when the approach outperforms or matches established baselines.
Calculate Your Potential ROI
Estimate the financial and operational benefits of integrating advanced control systems into your enterprise architecture.
Our Implementation Roadmap
A structured approach to integrating control systems into your computing infrastructure for predictable success.
Discovery & Strategy
In-depth assessment of current systems, identification of control points, and definition of performance objectives and constraints.
Modeling & Design
Development of dynamic models for computing resources and workloads, followed by control system design and simulation.
Prototyping & Integration
Building and testing initial prototypes, integrating control components into existing infrastructure, and iterative refinement.
Deployment & Optimization
Phased rollout, continuous monitoring, and adaptive optimization to ensure long-term stability and efficiency.
Ready to Transform Your Computing Systems?
Leverage advanced control theory to build more efficient, robust, and autonomous enterprise systems. Book a free consultation to discuss your specific challenges and opportunities.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat