Journal of Grid Computing 1: 1–2, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands.
1
Editors’ Message With the ever increasing amounts of data being generated by current grand challenge scientific applications, scientists are looking to Grid computing as a means for handling huge datasets. Grids are more than a technology for scientific computing, however. Potential business implications of the same technology are attracting major IT companies to enter the field. The Grid community and many others believe that the Grid has the potential to change the world even more than the Web has done in the past decade. Several test beds and experimental Grid systems have already demonstrated the power of the Grid. To reach the maturity of the Web, however, requires significant additional research and experimentation. A large number of research projects are under way – from the United States to Europe and as far as Japan and Australia. The rapidly expanding Grid community has established the Global Grid Forum (GGF) in order to standardize the major Grid protocols and to attack the major problems of Grid computing in a well-coordinated way. To date, however, the Grid community has been working without a journal that is devoted to this important new technology. Recognizing the need for and benefit of such a publication, and with the enthusiastic support of the Grid community, Kluwer has decided to launch a new journal. We proudly announce the inaugural issue of the Journal of Grid Computing. The journal is intended to provide a forum for exchanging information about developments in Grid computing. We are convinced that Grid computing is a rich research area with important new scientific results of interest to the whole community. For this first issue, we invited authors to discuss a wide range of research topics in Grid computing, from an overall definition of Grid computing to low-level concerns such as network monitoring. We also invited the chair of the Global Grid Forum to write about the most important aspects and goals of the GGF. We envision a close relationship between the journal and the Global Grid Forum. For future issues, we encourage every member of the Grid community to submit papers of the highest quality. The papers may involve purely theoretical issues, or may provide details about middleware or architecture, or may focus on an application in science or industry. The articles in this first issue convey some of the excitement of this promising new field as well as an idea of the wide range of activities in Grid computing. The authors of the first paper apply the Abstract State Machine (ASM) semantical approach to define the semantics of Grids and to express formally the main differences between traditional distributed computing and the novel Grid computing. Based on the runtime semantics of distributed systems, they systematically derive and prove the essential attributes that distinguish Grids from other distributed computing environments. They propose ASM as an alternative to the currently accepted models for determining whether a distributed system is a Grid. Motivating this proposal is the awareness that many systems are called Grid today simply because Grid is a fashionable term by which researchers can attract research grants and companies can promote their product on the market. Workflow creation and management are crucial issues in the Grid in order to solve the highly complex grand challenge problems. Most of the existing Grid systems require that the user produces the workflow either by a GUI or by a commandlike interface. The second paper proposes an alternative approach in which job workflows are automatically generated. The authors present two workflow generators. The first one maps an abstract workflow defined in terms of application-level components to the set of available Grid resources. The second generator, based on AI planning technologies, takes a wider perspective and not only performs the abstract to concrete mapping but also enables the construction of the abstract workflow based on the available components. The third paper addresses the problem of how to program the Grid, how to write Grid-enabled programs that can be easily and efficiently executed in a Grid environment. An exciting question concerning the creation of Gridenabled programs is whether we need revolutionary new programming paradigms to program the Grid or we can adapt existing programming models for the Grid. The authors vote for the latter approach and describe a GridRPC programming system, called Ninf-G, that realizes an RPC mechanism tailored for the Grid. Ninf-G is implemented on top of the Globus Toolkit, and the authors show preliminary performance evaluation in both WAN and LAN environments.
2 A complex problem in Grid computing today is scheduling computation and data in a data Grid where largescale data-intensive problems harness geographically distributed resources. Authors of the fourth paper propose a general and extensible scheduling architecture that can cope with the combined optimization of replication and scheduling strategies concerning the usual Grid parameters, such as resource utilization, response time, global and local allocation policies, and scalability. The authors simulated a wide range of scheduling strategies and parameters and obtained good results by scheduling jobs to locations that contain the needed data and asynchronously replicating popular datasets to remote sites. The final paper provides resource measurements across a wide range of network architectures, environments, and implementations. This explicit information about network resources can be used by applications to implement application-specific adaptation mechanisms. The authors describe the Remos Network Monitoring System, which is especially useful for applications that must make explicit configuration decisions, such as selecting a server from a set of candidates, selecting a set of compute nodes with certain connectivity properties, or deciding between local or remote execution. Remos collects appropriately detailed information at each site and distributes this information where needed in a scalable manner. Remos has been implemented and tested in a variety of networks and is in use in several different environments. With this selection of papers we believe that readers not only will learn about recent advances in Grid computing but also will be challenged to think about the Grid in new ways. We see the Journal of Grid Computing continuing this fine beginning, promoting understanding and discussion and advancing the field of Grid computing itself.
IAN FOSTER and PÉTER KACSUK Editors-in-chief