www'Pegasus'lse'ac'uk

1
www.Pegasus.lse.ac.uk
INFORMATION SYSTEMS AND INNOVATION
GROUP Department of Management
Distributed development of Grids Exploring the
collaborative practices of particle physicists as
they develop a Grid for the LHC
Research goals?
Introduction
R.Q How large-scale systems are developed in a
distributed/global way? We examine the systems
development practices of the particle physics
community as it develops its Grid.
With the globalization of the development
process the limitations of traditional systems
development are even more obvious. Literature
suggests that agile practices are not the silver
bullet to the challenges of global systems
development .
The community has a successful record in
developing cutting edge technology (e.g. the
Web!) and presents a context where distinctive
collaborative practices have emerged. Developmen
t of Grids poses new, underexplored opportunities
for understanding distributed collaborative
systems development. Within the particle
physics context, we therefore explore How
collaboration occurs The working practices
employed in this distributed development
endeavour
Global systems development demands new
development practices that focus on collaborative
ideas, since the nature of the problem and the
environment are now different. We here argue
that successful distributed development requires
not only flexibility/agility but also
structure/discipline together with creating a
sense of community and therefore construct an
identity for those involved in the development.
We are interested not just in the rationalistic
processes of design, but also in the emergent
practices, the improvisation, and bricolage.
Theory...
Some Findings...
As social researchers, we draw upon different
theories and concepts in an attempt to find ways
to analyze and explain our findings.
Activity theory the projects under study are
framed as a complex activity system influenced by
the communitys norms, culture, history and
collaborative practices. Literature global
systems development, open source, global
outsourcing, communities of practice.
Over 60 semi-structured interviews with members
of LCG, EGEE and GridPP as well as observation of
their major meetings and workshops. Findings
include It is more reacting to things that
go wrong, than actually planning on what you
should be doing. Here you have to be
much more agile. It doesnt mean that you should
drop the methods away, it just means you should
rotate them faster, and have smaller steps
In the past we had joint development of a group
of people who were distributed, but that was not
working. It was a painful processnow, we are
trying to go to a situation where one component
is being developed by the same group of people
who are all in one placethe idea here is to have
strong collaboration within the developers group
as well as collaboration with other countries in
order to share knowledge and expertise.
Collaboration and building community is really
important for distributed development. We work a
lot using mailing lists you can see the
different attitude people have before and after
they meet in person in those mailing lists.
Building some kind of a group is important one
way for this to happen is through consensus.
Contributions...
We aim to provide Practical
recommendations for those considering the
collaborative development of large-scale systems.
Suggestions for how such collaboration
might be aided. A recipe for distributed
systems development based on what physicists
perceive as the enablers - characteristics for
their success. A new form of systems
development practice.
Authors Avgousta Kyriakidou and Will Venters