The self-adaptation to dynamic failures for efficient virtualorganization formations in grid computing context

摘要

Grid computing aims to enable "resource sharing and coordinated problem solving indynamic, multi-institutional virtual organizations (VOs)". However, due to the nature ofheterogeneous and dynamic resources, dynamic failures in the distributed grid environ-ment usually occur more than in traditional computation platforms, which cause failedVO formations. In this paper, we develop a novel self-adaptive mechanism to dynamic fail-ures during VO formations. Such a self-adaptive scheme allows an individual and memberof VOs to automatically find other available or replaceable one once a failure happens andtherefore makes systems automatically recover from dynamic failures. We define dynamicfailure situations of a system by using two standard indicators: mean time between fail-ures (MTBF) and mean time to recover (MTTR). We model both MTBF and MTTR as Poissondistributions. We investigate and analyze the efficiency of the proposed self-adaptationmechanism to dynamic failures by comparing the success probability of VO formationsbefore and after adopting it in three different cases: (1) different failure situations; (2) dif-ferent organizational structures and scales; (3) different task complexities. The experimen-tal results show that the proposed scheme can automatically adapt to dynamic failures andeffectively improve the dynamic VO formation performance in the event of node failures,which provide a valuable addition to the field.