Implementing Distributed Server Groups for the World Wide Web.

Abstract

The World Wide Web (WWW) has recently become a very popular facility for the dissemination of information. As a result of this popularity, it is experiencing rapidly increasing traffic load. Single machine servers cannot keep pace with the ever greater load being placed upon them. To alleviate this problem, we have implemented a distributed Web server group. The server group can effectively balance request load amongst its members (within about 10% of optimal), and client response time is no worse than in the single server case. Client response time was not improved because the measured client traffic consumed all available network throughput. The distributed operation of the server groups is completely transparent to standard Web clients.

Supporting quality of service in HTTP servers

“…We will focus only on the distributed HTTP server work [15,14,41 because of the similarity in the issues addressed by these approaches and our approach. The focus in the distributed server research has been on using the resources of distributed hosts to increase the throughput of HTTP servers.…”

“…The research has, thus, focussed on developing techuiques (such as information caching [7,20,9,231 and distribution, partitioning [16] of server load across clients and servers, and parallelization [15,4,14, 181 of HTTP servers over SMPs and workstations) for eliminating performance bottlenecks arising due to the lack of sufficieut CPU, disk, and network Permission to make digital or herd copies of ail or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the fti page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.…”

Redirection algorithms for load sharing in distributed Web-server systems

“…Both the Cisco DistributedDirector (CiscoDD) [4] and the Distributed Server Groups (DSG) dispatcher [10] are based on a totally centralized approach. All the requests reaches the dispatcher that uses the HTTP redirection mechanism to redirect the requests to the most appropriate Web server.…”

“…All the requests reaches the dispatcher that uses the HTTP redirection mechanism to redirect the requests to the most appropriate Web server. The dispatcher of DSG selects the least loaded server [10], while that of CiscoDD uses some network metric to estimate the server with the closest proximity to the client that has originates a request [4]. Both DSG and CiscoDD use redirection as the only means to dispatch requests.…”

A gateway-assisted approach toward QoS adaptations

“…This proxy transcodes the server data so that it matches the client's reconfigured QoS needs, thus enabling the client to dynamically alter its execution level [2]. The second approach is to provide a gateway between the client and a series of servers [5]. In this approach, the gateway selects one server from a series of servers for each client, with each server providing an equivalent data stream.…”