Early Harvest:

Workshops for a first cutting of middleware issues in higher education

Summary

A common vision has emerged of a second level of national networking infrastructure. This layer of services provides mechanisms to allow users to securely and effectively access networked-based resources and readily collaborate across higher education. Such tools should be extensions of a core campus infrastructure that serves broad campus needs as well.

The intent of this proposal is to provide immediate velocity to the national efforts to deploy a core set of interoperable middleware services for research and instruction and inform a longer-term process to achieve a global middleware infrastructure. These goals will be achieved by gathering, integrating and distilling current wisdom from leading edge sites as roadmaps and convening a group of "study" institutions to both tune these "best practice" documents and begin local development of core middleware services. Further, a targeted but representative community of advanced application developers will be brought in to integrate their systems with these core middleware services. To inform longer-term middleware development processes, a report will be provided back to NSF on subsequent, larger-scale approaches to fostering this next generation of network services.

 

Background

The next stage of higher education networking infrastructure must shift from focusing on fiber optic circuits to investing in and creating a "services rich" environment to enable an increasing number of leading edge applications. Middleware is defined as common applications services where standardization in interfaces, tools, practices, and policy aids efficiency of applications development, interoperability, portability, and management. Middleware includes basic information technology functionalities and sophisticated production tools with which research communities could build their own collaboration environments. Clearly, these services must also be consistent with and benefit local campus IT infrastructure needs.

There are a cluster of middleware services that are closely associated with each other and appear to represent a core set that must be designed and implemented in concert. These areas include identifiers, authentication services, directories, and authorization. It is these specific territories that will be the focus of much of the work specified below. At the same time, such services need to be motivated and driven by scientific or instructional applications. This proposal also includes activities designed to couple this core middleware effectively with a set of specific applications being deployed by a distributed scientific community to both inform the middleware development and motivate campus deployments.

This area is of critical importance, both to the advance of scientific inquiry and to the use of technology by the broader community. Substantial amounts in both time and money will be needed to build a functional national middleware infrastructure for higher education and research. Early investments to begin the process and inform the future are certainly warranted.

 

Proposal

This proposal requests support for the following work:

1. Develop roadmaps for the key components of middleware.

The purpose of the roadmaps are to provide both CIO's and senior campus technical management with understanding of the issues that must be addressed in deploying the key components of a campus middleware infrastructure. The roadmaps will address both technical and policy issues. The roadmaps are being roughed out as a part of the existing I2MI effort, but grant resources will be used to assemble a group of campus experts for fuller development of the best practices.

Each roadmap will be highly structured, consisting of a broad CIO-level view of the key issues and design/implementation approaches. Within each issue will be links to greater technical discussions, terminating in pointers to specific products and papers. The roadmaps are intended to be multi-level living documents, providing overview information for CIO-types with links to deeper discussions and ultimately technical references and best practices. Roadmaps will include self-assessment surveys to help members evaluate their middleware infrastructures.

These reports will be a unique resource for several reasons. First, they will be designed for easy update and maintenance. This dynamic aspect will allow them to be living documents, and reflect the growing depth of knowledge about middleware. Secondly, their multilevel structure will permit them to be a resource for both CIO's and technical staff. Thirdly, their existence as a reference document may well serve to foster convergence of standards in practices. Lastly, unlike any other service out there, these roadmaps will be oriented around the distinctive aspects of higher education, such as mobility of client, loosely coupled domains, close cross-institutional interactions, FERPA and Open Records, etc. Identifying technical and policy couplings is critical to successful design of middleware in academia.

The roadmaps will be vetted within the higher ed community, and submitted to the IETF for inclusion in the RFC Best Practices series.

 

2. Create a study group of CIO's and associated campus IT architects to test and use roadmaps

It is important to get an early understanding of the issues that will be encountered by a variety of campuses in middleware deployment. This understanding will help shape and refine the roadmaps, improve the associated self-assessment tools, create some common practices, define additional requirements for interoperability, and move at least several institutions along the path towards a middleware infrastructure.

While much of the development of I2 Middleware will be driven by a few advanced schools, most of the challenges will be encountered by the broader I2 membership. A group of "not-the-usual-suspects" schools (approximately10-12) will be assembled to be trained on the issues, begin their deployments, and offer feedback to the overall initiative on tactical issues of campus middleware infrastructure.

In consultation with NSF and others, the selection of this study group will reflect demographic and geographic diversity. Participating institutions will be required to commit appropriate resources in time and money to begin their middleware deployments. While large-scale deployments on these campuses is unlikely in the time-frame of this grant, they are anticipated to have engaged the institutional processes and begun technical experimentation.

The study group will consist of a CIO and a lead IT architect from each selected campus. They will meet twice, once early to be informed by the draft roadmaps (and shape their form) and again six months later to enhance the roadmaps with their own experiences and information.

 

3. Integrate core middleware services with a set of advanced applications to demonstrate performance and motivate additional application development.

The middleware policy and technical development must work for the distributed scientific and computational environments that are emerging to support research. We need to explore how to gather requirements for specific applications and fit them into the broader context. And, we must understand how the requirements process can be efficiently applied to other discipline-specific applications. Therefore, we propose working with a deployed application that requires many of the middleware components we are addressing: the NSF-funded Space Physics and Aeronomy Research Collaboratory (http://geneva.crew.umich.edu/SPARC/). This is a web-based collaboration environment involving scientists around the U.S. and around the world. The NSF-funded research has focused on building the collaboration tools for scientists. This investigation will pursue security (identification, authentication, authorization), resource discovery (directory), multicast, and QoS issues for SPARC in the context of the facilities provided by campuses, the national networking infrastructure, and international cooperation (e.g., through STARTAP).

The nature of the work with SPARC would be twofold. First, a workshop would be held with SPARC's applications developers, SPARC end users, and middleware providers to define the middleware requirements of the application. Following the workshop, a set of detailed functional middleware specifications, including pointers to reference implementations, would be created. These implementations would include certificate and schema definitions and templates that enable the applications and can be included in appropriate campus and lab middleware deployments.

First, a workshop would be held with applications developers and some end users to define the middleware requirements of the applications and the nature of the environments that end users are in. Then a set of detailed functional middleware specifications, including reference implementations, would be created. These implementations would include certificate and schema definitions and templates that enable the applications and can be included in appropriate campus and lab middleware deployments.

As a result, the applications within SPARC could be more widely deployed; the motivation for campus deployment of middleware will be increased; and a model will be established for other applications areas to follow in definition and implementation.

 

4. Inform the NSF on specific services that need to be included in future efforts to foster middleware infrastructure.

It is clear that the lack of a widely deployed national interoperable middleware infrastructure is limiting progress in the use of high-performance networks and the development of advanced scientific applications. A series of workshops sponsored by NSF have indicated the need for NSF activities to support the development of middleware, but those efforts have generally not identified specific services that are required nationally or would provide leverage to campuses and research institutions in their deployments.

This project would provide NSF with a report that details the needs in this area. Recommendations would be drawn from both the CIO study group and from technical advisory groups associated with the roadmap development. In addition, the project offers to run a small series of seminars for directorate staff on the results of the early harvest workshops and how it could facilitate the advance of science in other disciplines.

 

Partnerships

The tasks of shaping and disseminating wisdom within higher education require broad participation. Function specifications for middleware services need to include multiple communities; broad dissemination is necessary both to foster wide experimentation and then convergence of standards. Therefore, the project team will include representation from key higher education organizations, including Educause, CNI, CREN, etc., as well as national advanced computing projects such as the PACI efforts.

 

Time Frames

Given the urgency in the need for middleware to support other advanced networking goals, Early Harvest proposes an accelerated schedule for its work. The following timeframe is, of course, dependent on many factors, including a quick response to the proposal.

August 1999

Develop detailed workshop plan and structure; select study group

September 1999

Host Early Harvest session

October 1999

Prepare rough drafts of roadmaps

October 1999

Present preliminary reports to I2 and Educause meetings

October 1999

Convene CIO study group and agree on processes and resources

November 1999

Meet with application developers at both SPARC and SC '99 to integrate middleware needs

December 1999

Finalize roadmaps, mount resources, begin broader dissemination

January 2000

Report to NSF and middleware issue seminars for NSF directorates

February 200

Update technical components of roadmaps

March 2000

Reconvene CIO study group to evaluate progress and enhance roadmaps

 

Qualifications of the Project Team

UCAID has been specifically created by a set of leading edge higher education institutions specifically to advance Internet development. These institutions have not only made significant financial commitments, but have agreed to pool intellectual resources to this goal. It has become evident throughout this community that the lack of an interoperable middleware infrastructure is the biggest impediment to wider-scale deployment of a next generation of network applications.

These institutions in general, and a subset of the most advanced institutions in particular, form the heart of the Glueworkers, a national working group assembled by UCAID to advise on middleware development in higher education.

The PI's have extensive experience in middleware and applications development in higher education, having served in a variety of management and technical leadership roles at major academic research institutions.