Text

Research Project 3.2

Dynamic deployment

Photo: https://unsplash.com/

Objective:

The main objective of RP3.2 is to develop innovative techniques for resource estimation and model-based analysis of executable models and intermediate code for computationally heavy work­ loads in heterogeneous computing environments. This includes core and edge platforms in the com­ puting continuum, focusing on critical resources like Worst-Case Execution Times (WCETs), response times, end-to-end delays, and energy consumption.

The project aims to analyse and estimate the resource requirements of workloads at both the code and software architecture levels. The code-level analysis involves examining the work­ loads' code for properties such as WCET and energy consumption. The software architecture analysis focuses on assessing software architecture models' response times and end-to-end delays. This dual­ level analysis is crucial for understanding and optimising the performance of workloads in complex computing.

The approach in the project involves two key strategies:

  1. Developing novel techniques for resource estimation at the code level, enabling precise analysis of workloads with respect to execution time and energy consumption.
  2. Conducting software architecture-level analysis to understand and optimise response times and end-to-end delays in the system.

Both strategies will utilise model-based analysis methods, offering a comprehensive view of resource utilisation in heterogeneous computing platforms. UC5 will be used by the project.

Expected outcomes:

  1. Development of a set of techniques and tools for effective resource estimation of workloads.
  2. Enhanced understanding and optimisation of worst-case execution times, response times, and energy consumption in heterogeneous computing environments.
  3. Improvement in workloads' dynamic allocation and performance efficiency, leading to better over- all system performance and resource utilisation in core-to-edge computing scenarios.

Staffing:Senior researchers (part of), 1 PhD-student funded by VR (part of).

Duration: 2024-2029

Partners