Václav Struhár

Autonomous Resource Mechanism for Real-time Containers

Student   Václav Struhár
Advisors   Moris Behnam
Alessandro V. Papadopoulos
Mohammad Ashjaei
Silviu S. Craciunas
Faculty Reviewer   Marcello Cinque, Università degli Studi di Napoli Federico II, Naples, Italy
Grading Committee   Valeria Cardellini, University of Rome Tor Vergata, Rome, Italy
Lei Feng, KTH, Stockholm, Sweden
Risat Pathan, Chalmers University of Technology, Gothenburg, Sweden
Kristina Lundqvist, Mälardalen University, Sweden (reserve)
Defence   Mälardalen University, Västerås, Sweden
Room Kappa and online (Link will be made public)
October 1, 2026 13:15
Abstract   Container-based virtualization has become an advantageous deployment model for heterogeneous distributed computing environments, including cloud and fog platforms, due to its lightweight isolation and near-native performance. By enabling efficient consolidation of physical resources, containers can reduce infrastructure footprint, lower operational costs, and improve system utilization. Despite these advantages, many industrial domains impose stringent domain-specific constraints, most notably temporal predictability (real-time behavior). Conventional container technologies do not inherently guarantee such predictability. In practice, containerized workloads often exhibit variability in execution and response times, driven by both their own resource demands and interference from co-located containers. This timing instability stems from contention over shared hardware resources, such as last-level caches (LLC), main memory, and data buses.

This doctoral thesis aims to enhance the temporal predictability of container-based virtualization by adjusting container resource reservations. It comprehensively investigates various aspects pertaining to soft real-time container-based virtualization. Specifically, the following contributions are presented: (i) a systematic literature review summarizes existing research directions aimed at enabling real-time container-based virtualization; (ii) a container orchestrator is designed to facilitate the deployment and adaptation of real-time containers with its implementation in Kubernetes; (iii) a Hierarchical Resource Orchestration Framework is developed for real-time containers, comprising a multi-layer hierarchy of resource management to facilitate resource adaptation; finally (iv) a virtualization of dynamic control systems is introduced that adapt the execution rates to system dynamics and optimize the overall performance.
Rules and Guidelines   The PhD procedure summary
Guidelines for Third-Cycle Studies at MDU
Thesis   Thesis
Included Papers   Paper A: Real-Time Containers: A Survey .
Paper B: REACT: Enabling Real-Time Container Orchestration .
Paper C: Resource Adaptation for Real-Time Containers Considering Quality of Control .
Paper D: Hierarchical Resource Orchestration Framework for Real-Time Containers .
Paper E: Container Orchestration in Edge Computing with Fluctuating Green Energy: A Multi-Armed Bandit Approach .
Paper F: RT-SCALER: Adaptive Resource Allocation Framework for Real-Time Containers .
Publications   Complete list of publications

Back to Research

Last modified: 2026-08-20 13:38:17 +0200