Geographically Distributed Software Development

Geographically distributed software development holds much promise for increasing market penetration and speeding up development cycles. However, it also comes with a set of new challenges for those developing the software, bought about by the distance among colleagues. This paper outlines a new research project underway to explore those issues and their implications for organizing g eographically distributed software development efforts. We also d escribe the a pproaches we a re taking towards providing solutions — in the form of processes and technology — to address the challenges of working remotely. INTRODUCTION Geographically distributed d evelopment has become a way of life for many divisions of Lucent Technologies, and all indications are that it will be increasingly prevalent i n the future. There are at least three compelling motivations. First, there is increased global demand for telecommunications products and services due to deregulation and innovations s uch as digital wireless technology. One c ondition some governments place on selling products and services in their country is that a ce rtain percentage of development resources be located in that country. Companies like Lucent are strongly encouraged to locate significant development resources in these countries. Second, there is a desire to speed up the development process and make it more efficient. With development locations in several countries widely spread across the globe, it becomes theoretically possible to d evelop around the clock by handing off work from one location to another. The potential cycle time reductions of such an arrangement are e xtremely attractive. Mastering distributed d evelopment also would create a large pool of resources, rather than many separate single-location pools, allowing resources to be used more efficiently by moving them among projects as needed. Third, there a re historical reasons for the geographic distribution of the development effort. The various components of the system itself have independent development histories. Some pieces were originally built in the United States, while others came from European standards. We believe that t his trend is s et t o continue, especially as companies invest in reuse efforts that encourage them to take what has been done, rather than rebuild. Despite the necessity, and p erhaps even desirability of geographically distributed development, it i s extremely difficult t o do successfully (see, e.g., [6]). The Lucent Technologies experience has probably been fairly typical, finding that geographic distribution introduces delays, misunderstandings, frustrations, and inefficiency. In response to this need, we have undertaken a research program in g eographically distributed product development with the following goals: • Create a portfolio of methods and techniques for addressin g communication and coordination issues in g eographically distributed collaboration. • Understand the various forms of collaborative software development. • Develop models that relate methods and techniques to the various forms of collaboration. • Understand the infrastructure demands. • Understand the role of cross-cultural communication barriers. PILOT ORGANIZATION We have c hosen GSM (Global System for Mobile c ommunication) Wireless development as our pilot because the project has s everal different kinds of collaborations across s ites involving v arious levels of “coupling.” [7]. Draft for review. Do not copy or quote. Draft for Review. Please do not copy or quote. 2 Some of their cross-site c ollaborations involve fairly clean divisions of labor and well-defined interfaces between the distributed groups. Other sites are tightly coupled; t hey exchange information frequently and make decisions that require constant synchronization. The GSM sites also span different languages and cultures, which makes the c oordination all t he more difficult. Finally, the organization is very important to the business and highly motivated to o vercome the substantial barriers it has encountered. By way of a brief background, GSM is the primary standard for cell phones in Europe. It is growing quite rapidly in nearly all areas of the world outside the US and Japan, and now has the largest global “market share” of all t he ce llular standards. Lucent Technologies currently has only a small share of the GSM market, but t he company has made significant i nvestment i n GSM development and has aggressive market goals. Product The infrastructure of a cellular network consists of three basic c omponents. The base station subsystem (BSS) is located in the field, with one or more a ntennas (one a t t he ce nter of each “cell”) and handles communication with handsets. Each BSS is connected to a mobile switching center ( MSC), the second major component, which is primarily a telephony switch that joins the BSSs' to the public switched telephone network, and o ften to d ata networks and even directly to some other MSCs as well. The third component i s the operations and maintenance ce nter ( OMC) which monitors the performance of the network and p lays a central role in maintaining it. Figure 1 shows a simplified version of a GSM wireless network.