Authors: Nasser Lotfi, Jamshid Tamouk
Keywords: Data Allocation Problem, Simulated Annealing, Variable Neighborhood Search.
Abstract
Data allocation problem (DAP) is of great importance in distributed database systems (DDS). Minimizing the total cost of
transactions and queries is the main objective of DAP which is
mostly affected by the volume of transmitting data through the
system. On the other hand, the volume of transmitting data
depends on the fragment-to-site allocations method. DAP as a
Np-hard problem has been widely solved by applying soft computing methods like evolutionary algorithms. In the continuation
of our previously published research, this paper proposes a novel
hybrid method based on Simulated Annealing Algorithm (SA)
and Variable Neighborhood Search (VNS) mechanism for Solving
DAP. To increase the performance, VNS mechanism is embedded into SA method in the proposed hybrid method. Technically
speaking, in order to discover more promising parts of search
space, the proposed method (VNSA) explores the search space
via SA and fulfills more exploitation by applying neighborhood
search mechanism. Moreover, due to the fact that both are a single solution-based method, they explore the search space faster
than population-based methods. Performance of the proposed
VNSA is experimentally evaluated using well-known benchmarks
reported in state-of-the-art literature, and evaluation outcomes
prove the robustness and fastness of the proposed hybrid method
(VNSA). Furthermore, the results exhibit that VNSA outperforms its competitors and achieves better results in majority of
test problems.
Nasser Lotfi
Faculty of Engineering, Cyprus Science University, Girne, N.
Cyprus via Mersin 10, Turkey
E-mail:
Jamshid Tamouk
Faculty of Engineering, Eastern Mediterranean University,
Famagusta, N. Cyprus via Mersin 10, Turkey
E-mail:
Fulltext
–
1.16 Mb