Browsing by Author "Ercetin, Ozgur"
Results Per Page
Sort Options
Item Market- Based Resource Allocation for Content Delivery in the Internet(2002) Ercetin, Ozgur; Tassiulas, Leandros; ISR; CSHCNThe Internet has evolved into a worldwide information backbone with vast user base. Internet users are diverse and they receive varying utilities from their network connections. In many cases high-level user decisions appear to have direct consequences on the network performance. In this thesis, our motivation is to characterize the effects of user behaviors in terms of lower layer network metrics such as network latency. We consider the Content Delivery Networks for our analysis, since they are the interconnection of network elements at the application layer and thus are directly affected by the user preferences.It has been a common practice to use caches to store the most popular data in order to improve user latency and reduce the network load. Recently, a more systematic approach to the caching has been developed in the framework of Content Delivery Networks. Content Delivery Networks (CDNs) are the networks of caches that are distributed throughout the Internet serving user requests directed to their subscriber web sites (publishers). They distribute the publisher's original content intelligently to the caching servers (surrogates) all over the world. Users receive their information from the surrogates, which are closer and usually much less loaded than the origin server. The objective is to minimize the user latency by intelligently distributing the content and serving the user requests from the most efficient surrogates. We use price-directed market based algorithms to achieve this goal. As it is the case in the current Internet, we model the agents with a selfish self-maximizing behavior and define the problem as a non-cooperative game played among the publishers and the surrogates. We show that the system has an equilibrium and if this equilibrium is unique, even though the agents are non-cooperative we achieve the global optimum. We also determine a uniqueness condition, which states that if the publishers are not willing to pay high amounts and the cache sizes are sufficiently small, then the equilibrium is unique. We noticed that the global system optimum that minimizes total average user latency requires the publishers to make very high investments, which in practice may prohibit the applicability of the distributed market method. Thus, we consider an investment strategy, which leads to a near-optimum system solution at much lower investments.
The abovementioned method gives publishers infinite granularity to determine their quality of service. Next, we investigated the case where the publishers can offer only finite number of QoS classes. In this model, surrogate partitions its total capacity among different service classes and among each class publishers receive equal shares of the resources. In our model, publishers try to get as large cache space as possible, while the surrogate is required to achieve fair allocation among the publishers. Specifically, each publisher should be charged the same if they receive equal share of caching space. We determine the optimal pricing strategy of the surrogate maximizing its revenue. We also analyzed the competition among two surrogates under this model and determined the condition that leads to a Nash equilibrium. We showed that at equilibrium surrogates peer as if they are a single combined surrogate server.
Item Next Generation Satellite Systems for Aeronautical Communications(2000) Ercetin, Ozgur; Ball, Michael O.; Tassiulas, Leandros; Tassiulas, Leandros; ISR; NEXTORThe US airspace is reaching its capacity with the current Air Traffic Control system and a number of flights that is constantly rising, and estimated to be over 54 million per year by 2002. The FAA has undertaken several projects to modernize the National Airspace System (NAS) to ensure the safety of the increasing number of flights. Of special importance is the modernization of the Air-Ground (A/G) Communications infrastructure, which is the heart of the air traffic control (ATC).The current plan in the modernization of the A/G communications is to migrate from analog voice only system to integrated digital voice and data system. The next generation satellite systems can be an alternative to the terrestrial A/G systems by their low propagation and transmission delays, global coverage, high capacity, and free flight suitable characteristics. In this paper, we give an overview of the current and the future ATC architectures, describe the systems and the communications issues in these systems, and develop a framework in which LEO/MEO next generation satellite systems can be integrated to the future ATC systems.
Item Optimal Cache Allocation Policies in Competitive Content Distribution Networks(2001) Ercetin, Ozgur; Tassiulas, Leandros; ISR; CSHCNExponential expansion in network dimensionality and user traffic has created substantial traffic congestion on the Internet. This congestion causes increased delays perceived by the users while downloading web pages. Users have considerably short patience, and when they do not start receiving information in a short while, they stop browsing the requested web page.As the commercial value of the Internet has become prevalent, the importance of keeping users at a site started to have direct translation into business value.
Proxy caching can alleviate problems caused by increased user traffic. In this paper, we consider the effects of real-world non-cooperative behavior of the network agents (servers and proxy caches) in overall network performance. Specifically, we consider a system where the proxy caches sell their caching space to the servers, and servers invest in these caches to provide lower latency to their users to keep them browsing their web pages and in turn to increase their revenues.
We determine optimal strategies of the agents that maximize their benefits. We show that such a system has an equilibrium point when no agent can increase its benefit by unilaterally updating its strategy. We show that under certain conditions this equilibrium leads to optimal cache allocation. We also show that an algorithm derived from this analysis is superior to currently implemented caching algorithms.
Item Push-Based Information Delivery in Two Stage Satellite-Terrestrial Systems(2000) Ercetin, Ozgur; Tassiulas, Leandros; ISR; CSHCNSatellite broadcast data delivery has inherent advantages in providing global access to information to everyone. However, users of satellite communications need expensive and cumbersome equipment to receive and transmit satellite signals. Furthermore, as the amount of information being broadcast increases, average user latency increases as well. In many situations, users in a locality may have similar interests and hence they can be better served by a local broadcast schedule. A two stage satellite-terrestrial wireless broadcast system can provide more efficient service. In such a system, main server broadcasts information via satellite to the geographically distributed local ground stations. Every station has limited buffer capacity to store the items broadcast by the satellite. According to their cache content, and the interests of their users, local stations deliver the information to their users via terrestrial wireless channel. We develop novel methods for the joint cache management and scheduling problem encountered in these systems. Our results demonstrate that two stage systems can provide more efficient data delivery compared to the single stage systems.