ABSTRACT
A bstract - The Internet has revolutionized the way computing is carried out by providing a new medium of information highway. For instance, with the emerging Internet computing infrastructures such as grid, peer-to-peer, and web services, it has become possible to deploy applications that support various Internet-wide collaborations. Yet, as Internet computing is becoming popular, it also presents a number of new challenges in coordinating resource usage. In particular, the resource allocation protocol that provides clear directives on the acquisition of shared resources, is complicated by the inherent characteristics of Internet computing, such as (i) the large-scale application deployment, (ii) the need for co-allocation of distributed resources that span multiple administrative domains, (iii) the lack of means for exchanging information between the inde pendent Internet applications, and finally (iv) the availability of functionally re dundant resources. The implication is that the resource allocation method must be decentralized, scalable, and most importantly it should be able to allocate the required resources with minimum time and cost. Recognizing the need for a systematic approach to the resource allocation for large-scale Internet comput ing, in this chapter we seek to promote the Internet resource allocation theory by presenting a new scalable resource allocation protocol that can address the characteristics of current Internet computing practices. The proposed protocol is free from deadlock and livelock, and seeks to effectively exploit the available alter native resource co-allocation schemes. Experimental results demonstrate that the proposed protocol yields a significant performance improvement over the existing deadlock prevention protocol.
