This paper presents a study of web content adaptation to improve server overload performance, as well as an implementation of a web content adaptation software prototype. When the request rate on a web server increases beyond server capacity, the server becomes overloaded and unresponsive. The TCP listen queue of the server's socket overflows exhibiting a drop-tail behavior. As a result, clients experience service outages. Since clients typically issue multiple requests over the duration of a session with the server, and since requests are dropped indiscriminately, all clients connecting to the server at overload are likely to experience connection failures, even though there may be enough capacity on the server to deliver all responses properly for a subset of clients. In this paper, we propose to resolve the overload problem by adapting delivered content to load conditions to alleviate overload. The premise is that successful delivery of a less resource intensive content under overlo...

We consider the problem of task assignment in a distributed system (such as a distributed Web server) in which task sizes are drawn from a heavy-tailed distribution. Many task assignment algorithms are based on the heuristic that balancing the load at the server hosts will result in optimal performance. We show this conventional wisdom is less true when the task size distribution is heavy-tailed (as is the case for Web file sizes). We introduce a new task assignment policy, called Size Interval Task Assignment with Variable Load (SITA-V). SITA-V purposely operates the server hosts at different loads, and directs smaller tasks to the lighter-loaded hosts.

This paper proposes a protocol suite for dynamic replication and migration of Internet objects. It consists of an algorithm for deciding on the number and location of object replicas and an algorithm for distributing requests among currently available replicas. Our approach attempts to place replicas in the vicinity of a majority of requests while ensuring at the same time that no servers are overloaded. The request distribution algorithm uses the same simple mechanism to take into account both server proximity and load, without actually knowing the latter. The replica placement algorithm executes autonomously on each node, without the knowledge of other object replicas in the system. The proposed algorithms rely on the information available in databases maintained by Internet routers. A simulation study using synthetic workloads and the network backbone of UUNET, one of the largest Internet service providers, shows that the proposed protocol is effective in eliminating hot spots and ...

As commercial interest in the Internet grows, more and more companies are o#ering the service of hosting and providing access to information that belongs to third-party information providers. In the future, successful hosting services may host millions of objects on thousands of servers deployed around the globe. To provide reasonable access performance to popular resources, these resources will have to be mirrored on multiple servers. In this paper, we identify some challenges due to the scale that a platform for such global services would face, and propose an architecture capable of handling this scale. The proposed architecture has no bottleneck points. A trace-driven simulation using an access trace from AT&T's hosting service shows very promising results for our approach. Keywords: Hosting service, scalable architecture, dynamic replication, migration. 1 Introduction As commercial interest in the Internet grows, more and more companies are o#ering hosting services i.e. ...

Abstract--Server replication improves the ability of a service to handle a large number of clients. One of the important fac-tors in the efficient utilization of replicated servers is the ability to direct client requests to the "best " server, according to some optimality criteria. In the anycasting communication paradigm, a sender communicates with a receiver chosen from an anycast group of equivalent receivers. As such, anycasting is well suited to the problem of directing clients to replicated servers. This paper examines the definition and support of the anycasting paradigm at the application layer, providing a service that uses an anycast resolver to map an anycast domain name and a selection criteria into an IP address. By realizing anycasting in the appli-cation layer, we achieve flexibility in the optimization criteria and ease the deployment of the service. As a case study, we examine the performance of our system for a key service: replicated web servers. To this end, we develop an approach for estimating the response time that a client will experi-ence when accessing given servers. Such information is maintained in the anycast resolver that clients query to obtain the identity of the server with the best estimated response time. Our performance collection technique combines server push with resolver probes to estimate the expected response time without undue overhead. Our experiments show that selecting a server using our architecture and estimation technique can improve the client response time by a factor of two over nearest server selection and by a factor of four over random server selection. Index Terms--Anycasting, replication, server selection. I.

With ever increasing Web traffic, a distributed multi-server Web site can provide scalability and flexibility to cope with growing client demands. Load balancing algorithms to spread the requests across multiple Web servers are crucial to achieve the scalability. Various domain name server (DNS) based schedulers have been proposed in the literature, mainly for multiple homogeneous servers. The presence of heterogeneous Web servers not only increases the complexity of the DNS scheduling problem, but also makes previously proposed algorithms for homogeneous distributed systems not directly applicable. This leads us to propose new policies, called adaptive TTL algorithms, that take into account of both the uneven distribution of client request rates and heterogeneity of Web servers to adaptively set the time-to-live (TTL) value for each address mapping request. Extensive simulation results show that these strategies are robust and effective in balancing load among geographically distribut...

This paper investigates load balancing strategies for clustered Alexandria digital library (ADL) servers. The ADL system, which provides on-line information searching and browsing of spatially-referenced materials through the World Wide Web, involves intensive database I/O and heterogeneous CPU activities. Clustering servers can improve the scalability of the ADL system in response to a large number of simultaneous access requests. One difficulty addressed is that clustered workstation nodes may be non-uniform in terms of CPU and I/O speeds. An optimization scheme is proposed in this paper to dynamically monitor the resource availability, use a low-cost communication strategy for updating load information among nodes, and schedule requests based on both I/O and computation load indices. Since the accurate cost estimation for processing database-searching requests is difficult, a sampling and prediction scheme is used to identify the relative efficiency of nodes for satisfying...

This paper studies the problem of balancing the demand for content in a peer-to-peer network across heterogeneous peer nodes that hold replicas of the content. Previous decentralized load balancing techniques in distributed systems base their decisions on periodic updates containing information about load or available capacity observed at the serving entities. We show that these techniques do not work well in the peer-to-peer context; either they do not address peer node heterogeneity, or they suffer from significant load oscillations which result in unutilized capacity. We propose a new decentralized algorithm, Max-Cap, based on the maximum inherent capacities of the replica nodes. We show that unlike previous algorithms, it is not tied to the timeliness or frequency of updates, and consequently requires significantly less update overhead. Yet, Max-Cap can handle the heterogeneity of a peer-to-peer environment without suffering from load oscillations.

Traditional techniques for a distributed web server design rely on manipulation of central resources, such as routers or DNS services, to distribute requests designated for a single IP address to multiple web servers. The goal of the Distributed Cooperative Web Server (DCWS) system development is to explore application-level techniques for distributing web content. We achieve this by dynamically manipulating the hyperlinks stored within the web documents themselves. The DCWS system effectively eliminates the bottleneck of centralized resources, while balancing the load among distributed web servers. DCWS servers may be located in different networks, or even different continents and still balance load effectively. DCWS system design is fully compatible with existing HTTP protocol semantics and existing web client software products. keywords: scalable web servers, document migration, load balancing. 1 Introduction With the explosive popularity of the internet and the world wide web (WWW...

The proliferation of the Internet is leading to high expectation on the fast turnaround time. Clients abandoning their connections due to excessive downloading delays translates directly to profit losses. Hence, minimizing the latency perceived by end-users has become the primary performance objective compared to more traditional issues, such as server utilization. The two promising techniques to improve the Internet responsiveness are caching and replication. In this paper we present an overview of recent research in replication. We begin by arguing on the important role of replication in decreasing client perceived response time and proceed by illustrating the main topics that affect its successful deployment on the Internet. We analyze and characterize existing research, providing taxonomies and classifications whenever possible. Our discussion reveals several open problems and research directions. 1