Assigning tasks in a 24-h software development model

Abstract

With the advent of globalization and the Internet, the concept of global software development is gaining ground. The global development model opens up the possibility of 24-h software development by effectively utilizing the time zone differences. To harness the potential of the 24-h software development model for reducing the overall development time, a key issue is the allocation of project tasks to the resources in the distributed team. In this paper, we examine this issue of task allocation in order to minimize the completion time of a project. We discuss a model for distributed team across time zones and propose a task allocation algorithm for the same. We apply the approach on tasks of a few synthetic projects and two real projects and show that there is a potential to reduce the project duration as well as improve the resource utilization through 24-h development.

Introduction

With the advent of globalization, many companies have started expanding their presence round the world, which has resulted in a single organization with multiple geographically distributed development centers (DCs). With the evolution of internet and other technologies like instant messaging, and audio and video conferencing, communication among these DCs has become better and cheaper.

Most software organizations have DCs in different regions of the world largely for business and manpower related reasons. This idea of having DCs globally has evolved mostly due to factors like cost advantage, availability of resources or human capital, and proximity to local market and conditions. Consequently, most current uses of DCs revolve around achieving the business benefit for which DCs are started. For example, one common use of multiple DCs is to assign different projects or development of different products to the various DCs. The main benefit of the DCs in this approach is that a higher volume of work can be done. That is, the DCs are primarily a source of manpower—the fact that they are located in different time zones is purely incidental.

Sometimes multiple DCs are involved in developing one product. The common model used here is to partition the product into relatively independent components and have the different DCs independently develop the various components. In this approach also DCs are primarily a source of manpower and there is no other benefit of having the DCs in different time zones.

Though current approaches use the DCs primarily to enhance manpower, having DCs globally distributed brings the possibility of a 24-h development model where the different time-zones of different locations can be leveraged to reduce the total development time of a project, through daily hand offs of work (Herbsleb and Moitra, 2001). For leveraging the geographical spread of the DCs to reduce the overall time for a project requires proper development approaches.

One example of effective leveraging of the potential of 24-h work day is in software maintenance. There are many instances of products being supported in India where the customer in USA logs a complaint at the end of the day and finds the complaint resolved next morning—the resolution takes place in India when it is night in USA. If the defect fixing team was also in the same time zone as the team that finds the defect, it would generally take one additional day for the user to get the software fixed. (We are assuming that fixing takes 1 day.) That is, by leveraging the DCs in two opposite time zones, problem logging and resolution cycle for the user reduces from 2 working days to 1 working day.

However, not much work has been done in leveraging the power of 24-h work day for reducing the cycle time of a development project. Clearly, leveraging teams that are spread across time zones will complicate task scheduling and project management. A global software development effort makes a project multi-site and multi-cultural, and introduces a new set of communication, technical, managerial, and coordination challenges (Carmel, 1999, Herbsleb and Moitra, 2001), which can have a negative effect on the project effort and schedule (Herbsleb and Grinter, 1999, Herbsleb et al., 2001, Mockus and Weiss, 2001). Many suggestions have been made to mitigate the problems due to communication difficulties (Carmel, 1999, Carmel and Agarwal, 2001, Herbsleb and Grinter, 1999). Tool support has also been suggested to alleviate some of the problems (Ebert and De Neve, 2001).

Even if communication and coordination difficulties can be solved satisfactorily through suitable tools and technologies, can distributed software development provide benefits, and if so, to what extent? In this paper we address this question of what potential benefits can global software development bring in reducing the project execution time. An understanding of potential benefits can help evaluate whether the benefits are worth the cost of such systems.

One approach to leveraging distributed teams for reducing the cycle time is to use the Timeboxing process model (Jalote et al., 2004a, Jalote et al., 2004b). The Timeboxing approach is an iterative development approach which leverages the DCs by dividing each iteration into different stages and having different groups execute the different stages in a pipelined manner. Substantial reduction in delivery time is possible after the first few iterations. Though this approach leverages the different teams provided by the DCs, it does not utilize the 24-h framework for reducing the schedule.

In this work we focus on how having groups in different time zones can be leveraged to build a 24-h software factory that can reduce the completion time of a project as work is being done round-the-clock.

We consider a 24-h software factory model, where multiple teams that are geographically distributed across different time zones work on the same project. Intuitively, since this model supports a 24-h operation, it should be possible to reduce the development time. However, the potential of improvement is heavily dependent upon the degree of inter-dependency between project tasks and nature of their various constraints. For example, if the tasks have fewer dependencies between them, then the development time depends primarily on the number of resources available, so a multi-site team provides no extra benefit over a single-site team of the same size. Clearly, to maximize the reduction in completion time through the 24-h model, task assignment needs to be done carefully, so that the project can be completed in minimum execution time while satisfying all the constraints.

In this paper, we first present our model for 24-h development and then present a task scheduling algorithm. The algorithm takes the task model for a project and the set of available resources as input and produces a task-schedule with minimal schedule length. We then apply the approach to the task graphs of a few synthetically generated projects, and the task graphs of two real projects. We show that not only there is a substantial reduction in the completion time of a project, there is an improvement in resource utilization as well.

For scheduling, the two main inputs are the project tasks and the available resources for executing the tasks. We present the project execution model that we assume in the next section and the resource model in Section 3. Then we present our task scheduling approach in Section 4, and discuss some properties of this 24-h development model in Section 5. Section 6 gives some experimental evaluation of the approach, which is followed by conclusions.