Current Content Distribution Networks (CDN) deployment requires heavy infrastructure investment since a large number of servers have to be deployed over a wide area. This paper proposes a new paradigm where future CDNs are to be deployed over a leased server and network infrastructure. While this paradigm shift introduces a new dimension of flexibility, it requires the resource provisioning and object placement problems to be jointly considered. This paper formulates this new optimization problem, and presents solution to it based on Lagrangian relaxation and greedy heuristics.

Caching and replication have emerged as the two primary techniques for reducing the delay experienced by end users when downloading web pages. Even though these techniques may benefit from each other, previous research work tends to focus on either one of them separately. In this paper we investigate the potential performance gains by using a CDN server both as a replicator and as a proxy server. We assume a common storage space for both techniques, and develop an analytical model that characterizes caching performance under various system parameters. Based on the models predictions, we can reason whether it is beneficial to reduce the caching space in order to allocate extra replicas. The resulting problem of finding which object replicas should be created where, given that any free space will be used for caching, is NP-complete. Therefore, we propose a hybrid heuristic algorithm (based on the greedy paradigm), in order to solve the combined replica placement and storage allocation problem. Our simulation results indicate that a simple LRU caching scheme can considerably improve the response time of HTTP requests, when utilized over a replication-based infrastructure. 1.

This paper proposes a non-cooperative game theoretical replica allocation technique (NCOR) to reduce user perceived Web access delays. NCOR uses distributed agents that because of their local knowledge act in a self-interested manner in order to enhance the performance of the servers that they represent. This can lead to some performance gains for some servers but has the potential to negatively impact the overall system’s performance. NCOR uses an effective cost model to guarantee the overall system performance gain despite the self-interested actions of these agents. With spontaneous and non-deterministic strategies, the system can exhibit Nash equilibrium. However, that may or may not guaranteed system-wide performance at a given time. Furthermore, their can be multiple Nash equilibria, making it difficult to decide which one is the best. Instead, we use the notion of pure Nash equilibrium, which if achieved is guaranteed to ensure stable optimal performance. Pure Nash equilibrium can be only achieved by deterministic strategies. In general, the existence of a pure Nash equilibrium is remarkably hard to achieve; however, we prove the existence of such an equilibrium in NCOR. Experimental comparisons with several non-game theoretical techniques reveal that NCOR maintains superior solution quality, in terms of lower communication cost and reduced execution time. 1.

Data replication is used extensively in wide-area distributed systems to achieve low data-access latency. Minimizing the cost of the resources used for replication is a key problem in these systems. The paper proposes a method to calculate lower bounds for the replication cost required to achieve certain QoS goals. We obtain bounds for the general case as well as for certain classes of replica placement heuristics. We observe that the cost of heuristics depends heavily on the workload and QoS goal. Based on these results, we discuss the inherent properties of heuristics that affect their cost and applicability to different environments.

In this paper we survey the field of Algorithmic Foundations of the Internet, which is a new area within theoretical computer science. We consider six sample topics that illustrate the techniques and challenges in this field. 1

The next generation of scientific experiments and studies are being carried out by large collaborations of researchers distributed around the world engaged in analysis of huge collections of data generated by scientific instruments. Grid computing has emerged as an enabler for such collaborations as it aids communities in sharing resources to achieve common objectives. Data Grids provide services for accessing, replicating and managing data collections in these collaborations. Applications used in such Grids are distributed data-intensive, that is, they access and process distributed datasets to generate results. These applications need to transparently and efficiently access distributed data and computational resources. This thesis investigates properties of data-intensive computing environments and presents a software framework and algorithms for mapping distributed data-oriented applications to Grid resources. The thesis discusses the key concepts behind Data Grids and compares them with other data sharing and distribution mechanisms such as content delivery networks, peer-to-peer networks and distributed databases. This thesis provides comprehensive taxonomies that

Though the WWW has come a long way since when it was monikered the World Wide Wait, it is still not reliable during heavy workload conditions. Overloads due to flash arrival of users or diurnal workload patterns are known to exponentially increase download times. More recently, online banks and portals have been the target of Distributed Denial-of-Service (DDoS) attacks, which send a deluge of requests and drive away the legitimate users. This dissertation proposes a web hosting architecture consisting of a grid of clusters, to provide high-performance in the presence of standard overload conditions as well as resilience during attacks. The architecture’s high-performance component is provided by a server selection framework, Wide-Area ReDirection (WARD), which efficiently multiplexes resources across the cluster grid. Traditional approaches assume that minimizing network hop count minimizes client latency. In contrast, WARD’s server selection algorithm for-wards requests to the server that minimizes the total of estimated network and server delays. WARD is better-suited to handling overload conditions in dynamic web con-

A brief overview of network algorithms In the rapidly growing world of Internet infrastructures, we face many challenging new algorithmic problems. These arise in part because the usual assumptions made in problems of this general type may no longer hold. For example, many typical questions deal with massive data sets and huge networks of prohibitively large sizes so that the (exact) number of nodes or edges is no longer a useful parameter. Instead, only partial information can be obtained. In addition, the network and the data are evolving dynamically. For example, the number of Internet hosts as of July 2005 topped 353 million and the number of Web pages indexed by large search engines now exceeds 8 billion [24]. Driven by the need of current technology and guided by the myriad of existing examples of large scale networks, the area of network algorithms has been rapidly expanding and evolving into a new and rich field. This field straddles both theoretical computer science and networking through the strong interplay between theory and practice. Furthermore, this field is closely connected to a variety of areas including graph theory, game theory, probability, statistics, physics, bioinformatics and the social sciences. Through this connection, a wealth of knowledge and methodology from diverse areas have been brought into play. As a result, new tools and insights have been developing and enriching

apport de recherche

apport de recherche