A 'standard task graph set' is proposed for fair evaluation of multiprocessor scheduling algorithms. Developers of multiprocessor scheduling algorithms usually evaluate them using randomly generated task graphs. This makes it difficult to compare the performance of algorithms developed in different research groups. To make it possible to evaluate algorithms under the same conditions so that their performances can be compared fairly, this paper proposes a standard task graph set covering many of the proposed task graph generation methods. This paper also evaluates as examples two heuristic algorithms (CP and CP/MISF), a practical sequential optimization algorithm (DF/IHS), and a practical parallel optimization algorithm (PDF/IHS) using the proposed standard task graph set. This set is available at http://www.kasahara. elec.waseda.ac.jp/schedule/.
- Multiprocessor scheduling
- Performance evaluation
- Task graph
ASJC Scopus subject areas
- Industrial and Manufacturing Engineering
- Management Science and Operations Research