SwinDeW-C: A Peer-to-Peer Based Cloud Workflow System

Abstract

Workflow systems are designed to support the process automation of large scale business and scientific applications. In recent years, many workflow systems have been deployed on high performance computing infrastructures such as cluster, peer-to-peer (p2p), and grid computing (Moore, 2004; Wang, Jie, & Chen, 2009; Yang, Liu, Chen, Lignier, & Jin, 2007). One of the driving forces is the increasing demand of large scale instance and data/computation intensive workflow applications (large scale workflow applications for short) which are common in both eBusiness and eScience application areas. Typical examples (will be detailed in Section 13.2.1) include such as the transaction intensive nation-wide insurance claim application process; the data and computation intensive pulsar searching process in Astrophysics. Generally speaking, instance intensive applications are those processes which need to be executed for a large number of times sequentially within a very short period or concurrently with a large number of instances (Liu, Chen, Yang, & Jin, 2008; Liu et al., 2010; Yang et al., 2008). Therefore, large scale workflow applications normally require the support of high performance computing infrastructures (e.g. advanced CPU units, large memory space and high speed network), especially when workflow activities are of data and computation intensive themselves. In the real world, to accommodate such a request, expensive computing infrastructures including such as supercomputers and data servers are bought, installed, integrated and maintained with huge cost by system users