Decentralizing execution of composite web services

  • Authors:

  • Mangala Gowri Nanda;Satish Chandra;Vivek Sarkar

  • Affiliations:

  • IBM India Research Laboratory;IBM India Research Laboratory;IBM T. J. Watson Research Center

  • Venue:
  • OOPSLA '04 Proceedings of the 19th annual ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Distributed enterprise applications today are increasingly being built from services available over the web. A unit of functionality in this framework is a web service, a software application that exposes a set of "typed'' connections that can be accessed over the web using standard protocols. These units can then be composed into a composite web service. BPEL (Business Process Execution Language) is a high-level distributed programming language for creating composite web services. Although a BPEL program invokes services distributed over several servers, the orchestration of these services is typically under centralized control. Because performance and throughput are major concerns in enterprise applications, it is important to remove the inefficiencies introduced by the centralized control. In a distributed, or decentralized orchestration, the BPEL program is partitioned into independent sub-programs that interact with each other without any centralized control. Decentralization can increase parallelism and reduce the amount of network traffic required for an application. This paper presents a technique to partition a composite web service written as a single BPEL program into an equivalent set of decentralized processes. It gives a new code partitioning algorithm to partition a BPEL program represented as a program dependence graph, with the goal of minimizing communication costs and maximizing the throughput of multiple concurrent instances of the input program. In contrast, much of the past work on dependence-based partitioning and scheduling seeks to minimize the completion time of a single instance of a program running in isolation. The paper also gives a cost model to estimate the throughput of a given code partition.