Self-assessment approach for resource management protocols in heterogeneous computational systems

TL;DR

This paper introduces a flexible heuristics-based self-assessment method for resource management in heterogeneous systems, enabling dynamic weighting, extensibility, and improved resource estimation for deployment decisions.

Contribution

It presents a novel, extensible heuristic algorithm that supports dynamic resource requirement weighting and can be integrated into various resource allocation protocols.

Findings

The approach provides accurate resource estimation.

It demonstrates scalability across different system sizes.

The method allows easy extension to new resource types.

Abstract

With an ever growing number of heterogeneous applicational services running on equally heterogeneous computational systems, the problem of resource management becomes more essential. Although current solutions consider some network and time requirements, they mostly handle a pre-defined list of resource types by design and, consequently, fail to provide an extensible solution to assess any other set of requirements or to switch strategies on its resource estimation. This work proposes an heuristics-based estimation solution to support any computational system as a self-assessment, including considerations on dynamically weighting the requirements, how to compute each node's capacity towards an admission request, and also offers the possibility to extend the list of resource types considered for assessment, which is an uncommon view in related works. This algorithm can be used by distributed and centralized resource allocation protocols to decide the best node(s) for a service intended for deployment. This approach was validated across its components and the results show that its performance is straightforward in resource estimation while allowing scalability and extensibility.