¦ Images of Virtuality –Athens Distributed large-scale systems development: Exploring the collaborative development of the particle physics Grid Avgousta Kyriakidou & Will Venters Information Systems and Innovation Group Department of Management The London School of Economics This research was undertaken as part of Pegasus EPSRC: Grant No: EP/D049954/1

¦ Images of Virtuality –Athens Overview  How large-scale systems are developed in a distributed way through a case study of a Grid development project – the LCG (Large hadron collider (LHC) Computing Grid).  Explore the collaborative practices employed by particle physicists, focus on collaboration, history & time.  Provide practical recommendations to those involved in globally distributed systems development.  Grid: Computers, storage devices and sensors connected by very fast networks – and Middleware to “virtualise” these resources. Grid computing for the LHC “demands” distributed systems development practices.  Activity as unit of analysis (minimal meaningful context for individual actions).

¦ Images of Virtuality –Athens Global Systems Development  Technological advances allow complex software development projects to run across geographically distributed environments (Oshri, Kotlarsky et al 2008).  GSD Complicates communication, coordination and control (Damian and Moitra 2006).  Limitations of traditional systems development and agile practices not the silver bullet to GSD.  GSD demands new development practices since the nature of the problem and the environment are now different.  How to ensure Agility of GSD practices?  What is the impact of the shift from traditional IT systems to information infrastructure such as Grids?

¦ Images of Virtuality –Athens Theoretical Framework  Activity Theory: how technology is collaboratively constructed to fulfil the objectives of a global community (Nardi 1996)  Present/frame the LCG project as a complex activity system influenced by the context, the community’s rules, norms and collaborative way of working.  Understand the Grid’s development as a series of contradictions between the elements of this activity system.  Activity Theory is driven by contradictions.:“systematic contradictions, manifested in disturbances and mundane innovations, offer possibilities for expansive developmental transformations” (Engestrom 2000).  Conceptualisation of collaborative breakdowns and tensions.

¦ Images of Virtuality –Athens Activity Theory Tools SubjectsObject RulesCommunityDivision of labour Transformation Process Desired Outcomes (Engestrom 1987) Unit of analysis – entire human activity in the collaborative context (Guy 2003). Activity is the context

¦ Images of Virtuality –Athens CASE: CERN  Demands globally distributed Grid of around 100,000 CPU(equiv).  Distributed due to funding and expertise availability.  A significant systems development challenge demanding distributed collaborative effort (Berman et al 2003).  PP well-known for its distributed nature (Chompalov et al 2002)  LCG cannot used plan based approach  Virtual meetings for development. Video-conf., s, blogs, wikis and travel.

¦ Images of Virtuality –Athens Developers’ Activity System  Desired Outcome: They are waiting for the LHC to produce data to analyse.  Object: Development of the LCG was realised when they understood that a Grid was necessary.  Formed global virtual alliances: “What physicists want to do cannot be done by a small group, its needs a large collaboration.”  Subjects: Global community of particle physicists and computer scientists.

¦ Images of Virtuality –Athens Developers’ Activity System  Tools: Activity is driven by the LHC.  Plan-based approach considered inappropriate for LCG.  No full overview of the system – modularised, fluid architecture  Draw on down-to-earth approaches embedded in PP tradition and history.  Pragmatic and improvisational in response to tight deadlines: “[PP] are more pragmatic in computing”. “they do [computing] just to resolve a problem for now”  Reactionary, flexible, adaptable.  “how can you possibly do a formal software engineering approach… if we are going to change it?”  Natural selection: hacked solutions.

¦ Images of Virtuality –Athens Developers’ Activity System  Community:  “Elite among sciences and among physics” (Traweek 1988).  Communitarian (Knorr-cetina 1999).  “we run big collaborations across every nation on earth and they always work”  “Set up to look like an experiment”  Shared goals and “sacred causes”.  Belief in “genius” nature of work – individual skill, competence and creativity.  Trust as central element.

¦ Images of Virtuality –Athens Developers’ Activity System  Rules: Collaboration is seen as inherent. Developing a strong sense of community.  Socialisation and Communication.  Improvising  Skills development via word of mouth.

¦ Images of Virtuality –Athens Developers Activity System  Division of Labour:  Volunteerism and flexible  “we don’t regard our roles as having fixed boundaries.”  “No strict hierarchy”  Politics is dispersed  “why was the web invented here? Because Tim had the freedom from hierarchy to spend a bit of time investigating something of interest to him.”

¦ Images of Virtuality –Athens Activity Theory Tools: SD Practices, Methodologies, Programming language Subjects: PP and Computer Scientists Object: Grid for LHC Rules: Collaboration, budgets Community: Collaboration, trust, pragmatism Division of labour: Fuzzy, Voluntary, lacking leadership Transformation: Collaborative SD Outcomes: Support for LHC

¦ Images of Virtuality –Athens Inner Contradictions… From the AT framework we observe contradictions:  Between physicists and computer scientists  Between developers and users  Between methodology and practice.  With the extreme size of the LHC experiments and previous experiments  Contradictions are arguably resolved through “expansive learning” – new technology, tools & work practices…

¦ Images of Virtuality –Athens New practices…  “Clusters of competence”  Balancing experimentation with discipline  A sense of belonging

¦ Images of Virtuality –Athens Conclusions  Lessons: 1.Structure the development effort in clusters of competence, 2.Encourage temporary co-location of developers, 3.Combine flexibility/agility with structure/discipline, 4.Create a sense of belonging and therefore construct identity for those involved in the development, 5.Facilitate human communication both through virtual means and face-to-face (at least every couple of months), 6.Create a trustworthy environment, 7.Have clear shared goals and rationale.

¦ Images of Virtuality –Athens Distributed large-scale systems development: Exploring the collaborative development of the particle physics Grid Avgousta Kyriakidou & Will Venters Information Systems and Innovation Group Department of Management The London School of Economics This research was undertaken as part of Pegasus EPSRC: Grant No: EP/D049954/1

¦ Images of Virtuality –Athens Why Activity theory? (Korpela, Mursu et al. 2002): AT can provide a theoretically founded but detailed and practicable procedure for studying systems development as a real-life collaborative work activity in context. (Kuutti 1996): AT is a powerful and clarifying descriptive tool, rather than a descriptive theory focusing on understanding and studying different forms of human activity and work practices as developmental processes with both individual and social level interlinked. ( Crawford and Hasan 2006): provides a useful paradigm for the ways in which human experience, needs and creativity shape the development and effectiveness of emerging technologies. (Leontiev (1978): Collectiveness and the notion of different actors sharing the same goals and constructing the same meanings are at the core of AT

¦ Images of Virtuality –Athens Why Activity theory (cont..) (Lehenkari and Hyysalo 2003): AT is particularly suitable to be used in studying contextually embedded interactions, as it contains features such as the recognition of actors, mediations, historicity, constructivity, dynamics etc.  helps the researcher to understand how these contextually embedded interactions influence the development and the systems development practices employed when developing the Grid (Engeström 1987): AT does not offer ready-made techniques and procedures for research, rather its conceptual tools must be concretized according to the specific nature of the object under study  Offers a place for innovation and creativity (Mwanza 2001): Activity notations - all possible interactions between the elements of the activity system. Questions that are specific to a particular combination can be then generated, and together with the notion of contradictions can guide the researcher’s field work and interpretation of findings.