From ESIWiki

Discussion Session I

What does your application need from the Grid? And how do you program it? (Thilo)

What makes an application a grid application? The first question being asked at this workshop should be about the requirements that applications have on "the grid". What is the added benefit that warrants all the hassle that comes with using grids?

After we have identified what applications can either get or would like to get from grids, the next question becomes: How do you program it? Or better: how would you like to program it? Given (early) application experience, which programming models, techniques, and styles are being used? And why? What are the good, the bad, and the ugly? What are the pain points that need to be resolved?

Based on the discussion so far, it would be ideal if we could identify a scope of applications to focus on in this theme on Distributed Programming Abstractions.

Discussion Session II

Grid Programming versus Programming Abstraction (Omer)

Various programming abstractions are currently available within the computer science community. These range from declarative programming (encompassing a range of styles from sub-routine oriented to object oriented), functional programming and logic programming styles. These styles have been used in various parallel computing approaches in the past -- such as the use of programming skeletons (in the context of functional programming), to the use of distributed object/subroutine based approaches centered on CORBA and MPI. These have also been employed within development tools such as Matlab, Mathematica and Netsolve (for instance). The aim of this session is to investigate which of these approaches are likely to be of benefit in developing Grid programs. In particular, the discussion will focus on why approaches based on functional and logic paradigms have not been exploited widely.

Discussion Session III

What Lessons Can We Learn from Parallel Programming? (Katz)

Many interesting reports have been written over the past 20 or so years discussing the issues that have prevented parallel computing from attracting large numbers of users and programmers. While the initial reports were fairly unique, later groups writing reports tended to rediscover that the same issues were still the main issues of concern. A seminal early workshop was the Pasadena Workshop on HPC Software Technology, and a good on-line reports are

The aim of this session is to understand what lessons we can learn from this history.

Discussion Session IV

Applications Revisited and the Road Ahead (Jha)

What we hope is going to be unique about this theme is the strong coupling between distributed programming models, techniques and abstraction on the one hand and real applications on the other. But what exactly are these applications? In some ways, at a very high level, one of the aims of the theme is to produce a document somewhat analogous to the "The Landscape of Parallel Computing Research: A View from Berkeley", but for Distributed Computing, i.e., survey the status of distributed applications in relation to the programming models and techniques, and in particular to identify analogous "dwarfs" for distributed computing as the Berkeley Group do for parallel computing.

This discussion session aims to ask the question, "What are the different types of Distributed Applications that this theme should investigate in detail and Why?".

This session will also try to summarize the two-day workshop, the main lessons learnt and discuss possible trajectories for the theme.

Return to Workshop I

DPA Discussion