I. Robert Chiang

Improving software team productivity

Choosing the most effective quantitative process design approach will not only improve a firms project management but, ultimately, the bottom line.

A Fault Threshold Policy to Manage Software Development Projects

This paper presents a project management policy in which the appearance of software faults during system construction is used to determine the timing of system integration activities (e.g., team meetings, analyzing modules for interface inconsistencies, system fault correction, and so on). System integration is performed only if a threshold fault count has been exceeded; otherwise, module development is allowed to continue. We derive an expression for calculating fault thresholds and analyze the policy to reveal the presence of three operating regions: (1) a region in which development should continue with no system integration, (2) a region in which system integration occurs if a threshold fault count has been exceeded, and (3) a region in which system integration should always take place. Analytical and numerical results demonstrate how the fault thresholds change with system complexity, team skill, development environment, and project schedule. We also show how learning that occurs during each round of system integration leads to less frequent integration in the future, and lower total construction effort. Simulation experiments reveal that the fault threshold policy can be applied even if several homogeneity assumptions in the model are relaxed, allowing for differences in the propensity among modules to accumulate faults and the effort needed to correct these faults. Finally, the fault threshold policy outperforms a fixed-release policy in which system integration occurs whenever a fixed number of modules has been released.

Strategic alignment and value maximization for IT project portfolios

Managing project portfolios has been a challenge to many IT organizations due to the size and complexity of their initiatives that are often cross-functional, fast changing, and transformational in nature. A governance process on project solicitation, evaluation, and monitoring is thus essential to ensure the resulting portfolio creates tangible values, balances across priorities, and supports business objectives. An optimization model to streamline the decision processes for IT portfolios and programs is proposed. We consider project characteristics such as the extent of strategic alignment, expected benefit, development cost, and cross-project synergy to maximize the portfolio value. We also consider team proficiency and resource availability to determine a project portfolio that could be implemented within the overall development time. The multi-objective model identifies the optimal mix among project types and the solution procedure efficiently produces recommendations that are superior to those found with current empirical techniques. We also describe an evolutionary algorithm to find approximate solutions to the optimization model. Possible extensions on how the optimization procedure can go beyond projects to also streamline decisions such as the renewal or replacement of in-flight applications is discussed.

Periodic cache replacement policy for dynamic content at application server

Web caching has been widely adopted to improve server responsiveness without resorting to costly infrastructure overhauls. However, due to the need to support real-time transactions and content customization, an increasing proportion of web pages contain fragments that are dynamically generated, typically results of database queries, rendering whole-page caching at browser or proxy level inappropriate. A promising approach has been to place dynamic fragments in a web application server cache for multiple web requests. In this paper, we propose a periodic cache replacement policy for dynamic web content at application server. Since the content of a dynamically generated web fragment could become stale, the value of a cached copy of the fragment will likely decrease as updates accumulate at the back-end database. Constantly replacing the cached copy of the fragment will increase its freshness and value, yet will also incur a high cache replacement cost. In addition, not all fragments are equally important to various applications and it is preferable to cache mission-critical fragments. The decision problem then consists of what fragments should be selected to cache and how frequently the cache should be replaced so that the total cache benefits per unit time is maximized. Numerical and simulation experiments show that the periodic cache replacement policy is robust and effective in handling dynamic content.

An Economic Analysis of Interconnection Arrangements Between Internet Backbone Providers

Transit and peering arrangements among Internet backbone providers (IBPs) are essential for the global delivery of communication services on the Internet. In addition, to support delay-sensitive applications (e.g., streaming and multimedia applications) it is important for IBPs to maintain high service quality even if the network is congested. One promising approach is to establish interconnection agreements among providers to dynamically trade network capacity. To make such interconnections possible in a competitive setting, we propose a pricing scheme that considers factors such as network utilization, link capacity, and the cost structure of the interconnecting participants. Our analyses show that the common sender keeps all (SKA) mode of settlement does not provide adequate incentives for collaboration; rather, the provider that delivers the packets should be suitably compensated at an equilibrium price. Two price equilibria are identified: The first favors slower IBPs, whereas the other is congestion based and can be more beneficial for faster IBPs. When cost asymmetries exist, the lower cost IBP needs to offer a price discount to induce participation. We show that a usage-based, utilization-adjusted interconnection agreement could align the costs and revenues of the providers while allowing them to meet more stringent quality of service requirements.

Delivery Consolidation and Service Competition Among Internet Service Providers

Abstract The infrastructure of the Internet, by and large, is maintained by Internet service providers (ISPs) that cater to regional customers and by Internet backbone providers (IBPs) that serve large organizations and ISPs. Some IBPs have recently branched into content delivery network (CDN) services; separately, other ISPs have started offering on-demand video streaming to compete with pure-play content providers. These developments have intensified the competition in both content delivery and media-streaming markets. For the content delivery market, we study the competition equilibriums by analyzing factors such as market share, cost structure, service pricing, and subscriber preference. Our approach helps identify conditions under which a content provider should choose an IBP over the incumbent CDN for content distribution. We also show how an IBP’s CDN venture affects its interconnection relationship with ISPs. For the streaming service market, we examine conditions under which a content provider would partner with an ISP to lower operating and marketing costs while providing a more streamlined subscriber experience. Analytically, our game-theoretical models can optimize key contracting and pricing strategies for multiple classes of service providers; empirically, insights derived from the proposed models have anticipated events that coincide with several recent developments in content delivery and streaming service markets.

Improving Web-Catalog Design for Easy Product Search

Building intuitive web sites is important for online businesses as positive experiences toward the virtual storefront could translate to customer goodwill and repeat visits. Streamlining web-site navigation is further motivated by the availability of comprehensive site-visit traces such as shopping carts and click-stream logs. We incorporate such sales and browsing patterns for the autonomous design of web catalogs by reducing the expected click count it takes to find related items. We first model catalog design as a task of placing items onto catalog pages, leading to a quadratic assignment formulation. For a more extensive redesign, such as when the number and location of links on each page are also decision variables, we propose a two-stage optimization procedure to ensure that click-count reduction is not achieved at the cost of excessive page cluttering. Our analysis reveals that an optimized catalog is robust against shifts in browsing behavior and that simplistic interfaces are preferred for users at either end of the experience spectrum. Using genetic algorithms, we find catalog designs that significantly outperform those found with a traditional greedy heuristic.

Supplier Involvement and Contract Design During New Product Development

Early supplier involvement (ESI) infuses upfront supplier resources and expertise to accelerate the research and development (R&D) timeline, and allow for risk sharing. Successful implementation of ESI in a new product development setting, however, remains elusive due to the intricacy of interfirm collaboration while dealing with unproven technology and market uncertainty. Extending from prior ESI studies on supplier selection, resource integration, and relationship management, we propose game theoretical contracting strategies to achieve manufacturer objectives, such as predictable design timelines, sufficient supplier commitment, and radical in-process innovations. Taking into account various project factors, such as revenue forecast, technical uncertainty, market competition, and team capability, we propose an incentive compatible mechanism based on real option analysis to suggest which project stage to best engage the supplier. The supplier, in turn, can follow our analysis to determine whether to participate, and if so, the appropriate level of resource commitment. The equilibrium analysis provides managerial insights into how to best balance the time-to-market mandate with the need for accruing significant innovations through supply chain partnerships.

Managing Software Development Process Using System Reliability Data

Due to the increasingly dynamic and complex nature of constructing software systems, the current-day system development is predominately incremental in nature. The decision of how long each development increment should last is critical to team productivity. This is because prolonged periods of coding without adequate system testing and team communication can result in considerable fault correction, while too much coordination can also be counterproductive by disrupting the progress of the development work. A major factor we consider that dictates the length of each development increment is the time-sensitive nature of the system reliability as a result of exponential software fault growth. Other factors include team communication overhead as well as the size effect incurred from the growing size of the system under construction. Our objective is to minimize the coordination overhead given that required system features are completed within the time allocated. Numerical results show how metrics collected from previous projects can be applied to the process design for future projects.