MemoirGRASP is a workflow execution strategy based on the GRASP (Greedy Randomized Adaptive Search Procedure) metaheuristic. It is designed to optimize the execution of data-intensive scientific workflows in multisite cloud environments by leveraging memoization—the reuse of intermediate results to avoid redundant computation.
Evaluates the trade-off between re-computation and storage costs to decide when cached data should be reused.
Uses a multi-start procedure (Construction + Local Search) to quickly find near-optimal scheduling plans.
Allows users to prioritize:
-
Makespan (
$\alpha_t$ ) -
Financial Cost (
$\alpha_b$ )
Demonstrated efficiency on real-world workflows such as:
- Montage (astronomy)
- Phenomenal (plant phenotyping)
All authors contributed substantially to the research objectives and methodology:
- Rodrigo A. P. Silva (UFF, Brazil) — Analysis and implementation
- Gaëtan Heidsieck (University of Göttingen, Germany) — Analysis and implementation
- Daniel de Oliveira (UFF, Brazil) — Supervision
- Yuri Frota (UFF, Brazil) — Supervision
- Esther Pacitti (Inria/LIRMM, France) — Supervision
- Patrick Valduriez (Inria/LNCC, France/Brazil) — Supervision
If you use MemoirGRASP in your research, please cite:
Silva, R. A. P., Heidsieck, G., Pacitti, E., Valduriez, P., Frota, Y., & de Oliveira, D. (2026).
Have We Seen These Data Before? A GRASP-based Execution Strategy for Cloud-based Workflows with Memoization.
Concurrency and Computation: Practice and Experience.
This research was supported by:
- HPDaSc (Inria-associated team with Brazil)
- Print/CAPES (nº 88887.310261/2018-00)
- CNPq (n° 145088/2019-7 and Universal nº 434421/2018-9)
- FAPERJ (n° E-26/202.806/2019)
Experimental data for this project is available at: