A SPQR-Tree Decomposition Technique for Efficient Cloud Inter-Domain Execution

Abstract

The challenge of workflow distribution has been recently addressed in the fields of grid computing, task delegation and cross-organizational communication, but little addressed for cross-zone application deployment in Cloud computing environments. This paper proposed a workflow decomposition technique to efficiently distribute a service workflow in various cloud domains. Our technique is based on SPQR fragment identification and considers a cost model that takes into account the activity execution cost on different domains, the limitations imposed by some activities related to their adequate execution domain and also the communication cost associated to connection links between Cloud Availability Zones. We compare our method with traditional SESE decomposition and BPEL transformation in an Amazon cloud environment, taking into account defined inter-domain patterns that cover most situations present in distributed workflow executions.