Resource Management and Budget Models – Computational Modeling

In Quality and Resource Management problems tradeoffs are explored between the resources allocated to an application and the quality provided by that application to a user. Such trade-offs are typically muli-objective and allow for different, Pareto optimal solutions. Models for QRM are often combined with a modular virtualization approach. Physical resources are shared between multiple application. A virtual platform is created by allocating resource budgets from each of the needed resources. Applications are then mapping onto such virtual platforms. Both platforms and applications are often defined in a modular way.

QRML is a model and a language that supports automation of online or offline quality and resource management. QRML is agnostic of the specifics or resources or applications. To effectively use it to optimize and configure virtual platforms, budget abstractions need to be defined that act as an interface between the promised resource availability of the resource and the requirement of the application.

For example, a processing budget could be abstractly modeled by a minimum number of cycles \(C\) in any real-time interval \(I\) of a given length. An application may be able to determine what throughput it can deliver based on that budget, for example using timed dataflow modelling and analysis methods, A TDMA based arbiter may be able to determine how many and which of its slots to allocate to the resource budget.

We have developed a prototype run-time resource management infrastructure based on these ideas and the the corresponding budget models, for the CompSOC platform [1, 2].

Selected Related Publications

S. T. Nikkhah, M. Geilen, D. Goswami, M. Koedam, A. Nelson, and K. Goossens, “A deployment framework for quality-sensitive applications in resource-constrained dynamic environments,” in 2021 24th euromicro conference on digital system design (dsd), 2021, pp. 212-220.
[Bibtex]

@INPROCEEDINGS{NGGea21,
author={Nikkhah, Shayan Tabatabaei and Geilen, Marc and Goswami, Dip and Koedam, Martijn and Nelson, Andrew and Goossens, Kees},
booktitle={2021 24th Euromicro Conference on Digital System Design (DSD)},
title={A Deployment Framework for Quality-Sensitive Applications in Resource-Constrained Dynamic Environments},
year={2021},
volume={},
number={},
pages={212-220},
doi={10.1109/DSD53832.2021.00042}
}

S. T. Nikkhah, M. Geilen, D. Goswami, and K. Goossens, “A performance analysis framework for real-time systems sharing multiple resources,” in 2020 design, automation & test in europe conference & exhibition (date), 2020, pp. 326-329.
[Bibtex]

@INPROCEEDINGS{NGGG20,
author={Nikkhah, Shayan Tabatabaei and Geilen, Marc and Goswami, Dip and Goossens, Kees},
booktitle={2020 Design, Automation & Test in Europe Conference & Exhibition (DATE)},
title={A Performance Analysis Framework for Real-Time Systems Sharing Multiple Resources},
year={2020},
volume={},
number={},
pages={326-329},
doi={10.23919/DATE48585.2020.9116514}
}

References

[1]

@INPROCEEDINGS{NGGea21,
author={Nikkhah, Shayan Tabatabaei and Geilen, Marc and Goswami, Dip and Koedam, Martijn and Nelson, Andrew and Goossens, Kees},
booktitle={2021 24th Euromicro Conference on Digital System Design (DSD)},
title={A Deployment Framework for Quality-Sensitive Applications in Resource-Constrained Dynamic Environments},
year={2021},
volume={},
number={},
pages={212-220},
doi={10.1109/DSD53832.2021.00042}
}

[2]

@INPROCEEDINGS{NGGG20,
author={Nikkhah, Shayan Tabatabaei and Geilen, Marc and Goswami, Dip and Goossens, Kees},
booktitle={2020 Design, Automation & Test in Europe Conference & Exhibition (DATE)},
title={A Performance Analysis Framework for Real-Time Systems Sharing Multiple Resources},
year={2020},
volume={},
number={},
pages={326-329},
doi={10.23919/DATE48585.2020.9116514}
}