Resource Allocation for Query Optimization in Data Grid Systems: Static Load Balancing Strategies

Shaoyi Yin (1) , Igor Epimakhov (2) , Franck Morvan (2) , Abdelkader Hameurlain (2)

1 IRIT - Institut de recherche en informatique de Toulouse
2 IRIT-PYRAMIDE - Optimisation Dynamique de Requêtes Réparties à grande échelle

Resource allocation is one of the principal stages of relational query processing in data grid systems. Static allocation methods allocate nodes to relational operations during query compilation. Existing heuristics did not take into account the multi-queries environment, where some nodes may become overloaded because they are allocated to too many concurrent queries. Dynamic resource allocation mechanisms are currently developed to modify the physical plan during query execution. In fact, when a node is detected to be overloaded, some of the operations on it will migrate. However, if the resource contention is too heavy in the initial execution plan, the operation migration cost may be very high. In this paper, we propose two load balancing strategies adopted during the static resource allocation phase, so that the workload is balanced at the beginning, the operation migration cost is decreased during the query execution, and therefore the average response time is reduced.

Mots clés: Load balancing; Data grid systems; Query optimization; Resource allocation

Domaines: Théorie de l'information [cs.IT]; Recherche d'information [cs.IR]

Shaoyi Yin, Igor Epimakhov, Franck Morvan, Abdelkader Hameurlain. Resource Allocation for Query Optimization in Data Grid Systems: Static Load Balancing Strategies. 17th East-European Conference on Advances in Databases and Information Systems (ADBIS 2013), Sep 2013, Genoa, Italy. pp.316-329. ⟨hal-01239717⟩