The Internet suffers from insufficient service quality to reliably support many real-time, high bitrate applications. This shortcoming is due in large part to users or providers not considering the effect their choices have on others ? in other words the resulting externalities. For example, users who download movies during the work day reduce the quality of the network for more time-critical applications; in a non-neutral network, individual ISPs may overcharge high-value applications if they do not consider the interactions of such charges across ISPs. Depending on the circumstances, the problem can be due to a combination of the following three effects: i) the technology of the network does not enable the user or provider to make better choices; ii) the network does not give the user or provider the information to make better choices; iii) the user or provider lacks an incentive to make better choices even if he can. This project will investigate how these effects impact service quality in the network. Using this analysis as a basis, remedies for these effects will be investigated. The remedies may include combinations of new protocols and technology, pricing mechanisms, and regulation ? including requirements and/or mechanisms for information disclosure.
Intellectual Merit: This project addresses fundamental scientific and engineering questions in networking. The scientific questions include suitable formulation and analysis of the strategic choices that the users, content providers, service providers, and other parties in the network face when they interact. The properties of the resulting equilibrium strategies and their distributed computation raise novel mathematical and algorithmic questions of broad relevance. Other fundamental scientific issues addressed include how to provide incentives for network evolution and for improved reliability. Engineering questions concern the scalability of the proposed algorithms and protocols, their security and robustness, how they can be deployed incrementally, and their extensibility as new technologies, services, and applications get implemented.
Broader Impact: The investigators expect that the work will contribute to a paradigm shift in the design of the future network. The focus on externalities, which is central to our project, has been largely unexplored and our preliminary investigations demonstrate its substantial importance. The investigations undertaken in the course of this project will suggest how network design should be modified to account for externalities and network informational imperfections. This project, with its focus on the intersection of the engineering and economic issues of network services, will offer ideal research experience to the project?s Ph.D. students. It is expected that the experiences and results from this project will greatly benefit the PI?s courses as well, in particular by making students aware of the inherent tradeoffs in engineering systems that function in a market context.
PROJECT OUTCOMES Jean Walrand, Venkat Anantharam, University of California, Berkeley John Musacchio, University of California, Santa Cruz Shyam Parekh, Alcatel-Lucent The internet has undoubtedly been a tremendous success – enabling new applications that have had a transformational impact on commerce and communication. Despite this success, the internet still suffers from inconsistent service quality for some real-time applications like video conferencing, and it also has been plagued by a growing number of security problems. The internet connects vast numbers of people into a system and consequently the quality and security of a user depends strongly on the choices of others in the Internet. For instance, your critical video conference quality is influenced by your neighbor who is congesting the same network downloading games, and your security is compromised when friends who have you on their email contact lists do not update their software with the latest security patches. In economics these couplings of one’s decisions on other’s well being are called externalities. Over the course of our project, we have studied the role of externalities in the internet, specifically how externalities degrade the welfare of the internet’s users. We also studied possible solutions to these problems. For example we have studied traffic congestion in wireless networks, and studied how an incentive scheme based on raffles might help distribute the traffic more evenly throughout the day, consequently making users better off. We have studied the question of net-neutrality, a notion that has many different dimensions but generally has to do with the question of whether ISPs can discriminate traffic from particular sources and impose charges or different service quality. In particular, we have investigated the effect of having neutral or non-neutral network on the investment and market-entry incentives of ISPs and content providers. We also have studied the tradeoff of allowing a fixed percentage of the network to be non-neutral and the rest neutral, rather than going all one way or the other. Related to that is the question of how public spectrum should be used – whether the license ought to be sold to a single provider or kept open like a public commons for everyone to use. We have developed an auction scheme that tried to find the best tradeoff between these two outcomes. Another area where externalities are important are in the decision of ISPs to upgrade to more advanced technologies. In some cases, being first to upgrade may confer a competitive advantage. In other cases, being first to upgrade to a technology that others are not using yet can be a big disadvantage since the technology may need a large fraction of the network using it to see its full benefit. We study this phenomenon and identify possible outcomes. Auctions for advertisements associated with key word searches on search engines are a major source of revenue in the Internet ecosystem. There are concerns that these auctions are susceptible for fraud, since the search engine gets paid when the ads are clicked on, and some of these clicks might be fake. We study this market and look at the incentives for search engines to invest in click-fraud detection capability, and we show the incentives are stronger for investment than one might expect. In the security realm, we have studied the effects of individuals making selfish choices of security investments (e.g. buying antivirus software or investing the time to update their computer) when there are externalities. We have also investigated the possibility of an insurance market for computer security, and studied the challenge of operating an insurer for a market where the risks are not as well known as in other domains like car and health insurance that have vastly more actuarial data.
