Building the Grid: An Integrated Services and Toolkit Architecture for Next Generation Networked Applications

By Ian Foster (foster@mcs.anl.gov)

Abstract:
A new class of advanced network-based applications is emerging, distinguished from today's Web browsers and other standard Internet applications by a coordinated use of not only networks but also endsystem computers, data archives, various sensors, and advanced human computer interfaces. These applications require services not provided by today's Internet and Web: they need a "Grid" that both integrates new types of resource into the network fabric and provides enhanced "middleware" services such as quality of service within the fabric itself.

Various groups within the research community and commercial sector are investigating these advanced applications. Successful large-scale experiments have been conducted and much has been learned about requirements for "middleware." Indeed, various federal agencies have started projects to create a Grid infrastructure: e.g., the National Science Foundation's Partnerships in Advanced Computational Infrastructure, NASA's Information Power Grid, and DOE's Next Generation Internet program.

With these and other efforts emerging, we believe that there is much to be gained from the definition of a broadly based Integrated Grid Architecture that can serve to guide the research, development, and deployment activities of the emerging Grid community. We argue that the definition of such an architecture will advance the Grid agenda by enabling the broad deployment and adoption of fundamental basic services such as security and network quality of service, and sharing of code across different applications with common requirements.

In this document, we develop a set of requirements for an Integrated Grid Architecture and present a candidate structure for this architecture.

The Grid: International Efforts in Global Computing

Mark Baker, Rajkumar Buyya and Domenico Laforenza

Abstract
The last decade has seen a considerable increase in commodity computer and network performance, mainly as a result of faster hardware and more sophisticated software. Nevertheless, there are still problems, in the fields of science, engineering and business, which cannot be dealt effectively with the current generation of supercomputers. In fact, due to their size and complexity, these problems are often numerically and/or data intensive and require a variety of heterogeneous resources that are not available from a single machine. A number of teams have conducted experimental studies on the cooperative use of geographically distributed resources conceived as a single powerful computer.

This new approach is known by several names, such as, metacomputing, seamless scalable computing, global computing, and more recently grid computing. The early efforts in grid computing started as a project to link supercomputing sites, but now it has grown far beyond its original intent. In fact, there are many applications that can benefit from the grid infrastructure, including collaborative engineering, data exploration, high throughput computing, and of course distributed supercomputing. Moreover, the rapid and impressive growth of the Internet, there has been a rising interest in web-based parallel computing. In fact, many projected have been incepted to exploit the Web as an infrastructure for running coarse-grained distributed parallel applications. In this context, the web has the capability to become a suitable and potentially infinite scalable metacomputer for parallel and collaborative work as well as a key technology to create a pervasive and ubiquitous grid infrastructure.

This paper aims to present the state-of-the-art of grid computing and attemppts to survey the major international adventures in developing this upcoming technology.

Indiana Portal Effort

By: Dennis Gannon, Randall Bramley, Juan Villacis, Venkatesh Choppella, Madhu Govindaraju, Nirmal Mukhi, Benjamin Temko, Rahul Indurkar, Sriram Krishnan, Ken Chiu

Abstract:
Powerpoint presentation that includes an overview of the "Standard" Portal Model, and Indiana University extensions for application & experiment management, and covers the following topics: Browser based "Notebook database"; Script Editor and Execution Environment; Component Proxies and component linking; File Management Issues.

Portals for Web Based Education and Computational Science

By Geoffrey C Fox (gcf@cs.fsu.edu)

Abstract:
We discuss the characteristics of portals defined as web-based interfaces to applications. In particular we focus on portals for either education or computing which we assume will be based on technologies developed for areas such as e-commerce and the large Enterprise Information Portal market. Interoperable portals should be based on interface standards, which are essentially hierarchical frameworks in the Java approach but are probably best defined in XML. We describe the underlying multi-tier architecture, the key architecture features of portals and give detailed descriptions for some computational science portals or problem solving environments. We suggest that one needs to introduce two object interfaces - termed resourceML and portal in XML. We show how this fits with recent education standards and describe web-based education in terms of collaborative portals.

Universal Design and The Grid

By Al Gilman

Abstract:
Trace Center has an EOT-PACI activity in the area of Universal Design for Advanced Computational Infrastructure. This document collects information based on our experiences in Universal Design that may be of value to the Grid Computing Environments Working Group or more widely in the Grid Forum.

Universal Design posits that products and services should be designed for the widest possible diversity of users, particularly as it affect demands on user capability. Likewise, that the remedy applied to accommodate special needs should be, and can often be if you look for it, a design feature yielding benefits across a far wider user base. A functional limitation is not a disability until there is something you can't do because of the functional limitation. Universal design aims to prevent functional limitations from turning into dis-abilities. And it aims to capitalize on the added depth of understanding of HCI issues that can come from looking at the design from a wider range of viewpoints.

The Grid is an attempt expand the possibilties for human benefit from distributed and remote computation by dint of an architecture of common practices. The possibilities are great, given the radical advances in both computer and network performance that are coming into use. This potential could be an engine for making information services more broadly available and accessible to a wider range of people, in particular people with disabilities. But this outcome requires attention to certain design issues early in the process and from the ground up in the architecture.

This whitepaper collects lessons learned from the area of Disability Access to Information Technology and Telecommunications. For example, one of the lessons learned is that the requirements of access by people with disability and access from small mobile devices are very similar. Thus paying attention to scalability issues in the characterization of services and resources will help people with screen readers, but not just people with screen readers. It will also help the Grid reach and be used by people via a wide variety of internet-interoperable devices, including mobile devices. The paper also outlines, in a provisional fashion for discussion, the relevance of these lessons learned to elements of the Grid architecture, starting at the Human/Computer Interface found in Grid User Environments.

Grid Programming Environments: A Component-Compatible Approach

Abstract: (note: this is extracted from the document Overview)
The following Grid Software-layer Model is often presented to help one understand the various roles in the overall programming environment. However, Item 3 is not typically included.

1. Applications
2. Problem Solving Environments (frameworks, toolkits, compilers, shells)
3. Programming Support Software
4. Distributed Computing Services/Grid Architecture
5. Networks of Resources and Operating System Software

The remaining discussion will have two goals. First the available component-based or component-like systems and standards will be highlighted. Then, a proposed approach will be presented that attempts to meet the needs described above. This proposal is not meant to be a complete design for a system. It is just an outline of an approach to stimulate a focused discussion.

Development of Web Toolkits for Computational Science Portals

Mary Thomas, Steve Mock, Jay Boisseau

Abstract
The NPACI HotPage is a user portal that provides views of a distributed set of HPC resources as either an integrated meta-system, or as individual machines. These web pages run on any web browser, regardless of system or geographical location and are supported by secure, encrypted login sessions where authenticated users can access their HPC system accounts and perform basic computational tasks.

We describe the development of the Grid Portals Toolkit (GridPort), which is based on the architecture developed for the HotPage, and will provide computational scientists and application developers with a set of simple, modular services and tools that allow application level, customized science portals development and facilitates seamless web-based access to distributed compute resources and grid services.

Mississippi Computational Web Portal

Tomasz Haupt

We are developing a domain specific Problem Solving Environment, or a Simulation Workspace, where the user can state complex multi-step problems, allocate all resources needed to solve them, and analyze results. In the proposed environment, definitions of problems, methods of solving them, and their solutions are persistently stored and, consequently, viewed and reused at later time, shared between researchers and engineers, and transitioned for operational or educational use. The Simulation Workspace is also an environment that extends the user desktop by providing a seamless access to remote computational resources (hardware, software, and data), hiding from the user complexity of a heterogeneous, distributed, high performance back end.

We envision the system as a shared collection of persistent components (representing applications, data, and tools) and services implemented using Enterprise JavaBeans (EJB) technology. The components can be organized into arbitrary complex hierarchies (referred to as contexts). This way we can build a multi-user system, where each user can customize his or her environment, the user can organize his or her work into independent projects, and most importantly, the user can compose a complex task by adding multiple applications ("atomic tasks") into a task context. The relationship between applications (such as data-flow dependency) is defined by the task descriptor, an XML document. The atomic tasks are PROXIES of actual codes and data, and as such they are defined in the platform independent fashion. For example, an application proxy possesses information on how to submit the corresponding code on the selected target machine (either the user’s desktop or a HPC back-end system). This information is entered to the system through a registration process. This way we can easily integrate both legacy and user applications into the PSE as well as separate user computational task specification from the resources allocation. The task contexts are persistent, and can be archived, cloned, and shared between users. The architecture of the proposed system allows for easy conversion of the PSE into a computational Web Portal, that is, building the front-end within ubiquitous and familiar Web browser interface.

The front-end of the PSE through an intuitive GUI will provide visual tools to compose (or reuse) the complex tasks, manage their execution, and control the flow of data for each step within the simulation pipeline. A wizard type user interface will help user learn the system and train users new to simulation methodology on the process without encumbering the advanced user, while automating and hiding the unnecessary. In addition, the user interface will provide access to local resources, such as grid generators and visualization packages.