This paper deals with data management in computer systems in which the computing nodes are connected by a relatively sparse network. We consider the problem of placing and accessing a set of shared objects that are read and written from the nodes in the network. These objects are, e.g., global variables in a parallel program, pages or cache lines in a virtual shared memory system, shared files in a distributed file system, or pages in the World Wide Web. A data management strategy consists of a placement strategy that maps the objects (possibly dynamically and with redundancy) to the nodes, and an access strategy that describes how reads and writes are handled by the system (including the routing). We investigate static and dynamic data management strategies. In the static model, we assume that we are given an application for which the rates of read and write accesses for all node--object pairs are known. The goal is to calculate a static placement of the objects to the nodes in the ne...

We present improved competitive on-line algorithms for the page replication problem and concentrate on important network topologies for which algorithms with a constant competitive ratio can be given. We develop an optimal randomized on-line replication algorithm for trees and uniform networks; its competitive ratio is approximately 1.58. This performance holds against oblivious adversaries. We also give a randomized memoryless replication algorithm for trees and uniform networks that is 2-competitive against adaptive on-line adversaries. Furthermore we consider on-line replication algorithms for rings and present general techniques that transform c-competitive algorithms for trees into 2c-competitive algorithms for rings. As a result we obtain a randomized on-line algorithm for rings that is 3.16-competitive. We also derive two 4-competitive on-line algorithms for rings which are either deterministic or randomized and memoryless. Again, the randomized results hold against oblivious ad...

This paper deals with static data management in computer systems connected by networks. A basic functionality in these systems is the interactive use of shared data objects that can be accessed from each computer in the system. Examples for these objects are les in distributed le systems, cache lines in virtual shared memory systems, or pages in the WWW. In the static scenario we are given read and write request frequencies for each computer-object pair. The goal is to calculate a placement of the objects to the memory modules, possibly with redundancy, such that a given cost function is minimized.

We define a novel measure of competitive performance for distributed algorithms based on throughput, the number of tasks that an algorithm can carry out in a fixed amount of work. This new measure complements the latency measure of Ajtai et al. [3], which measures how quickly an algorithm can finish tasks that start at specified times. An important property of the throughput measure is that it is modular: we define a notion of relative competitiveness with the property that a k-relatively competitive implementation of an object T using a subroutine U , combined with an l-competitive implementation of U , gives a kl-competitive algorithm for ...

Consider an arbitrary distributed network in which large numbers of objects are continuously being created, replicated, and destroyed. A basic problem arising in such an environment is that of organizing a distributed directory service for locating object copies. In this paper, we present a new data tracking scheme for locating nearby copies of objects in arbitrary distributed environments. Our tracking scheme supports ecient accesses to data objects while keeping the local memory overhead low. In particular, our tracking scheme achieves an expected polylog(n)- approximation in the cost of any access operation, for an arbitrary network. The memory overhead incurred by our scheme is O(polylog(n)) times the maximum number of objects stored at any node, with high probability. We also show that our tracking scheme adapts well to dynamic changes in the network. 1 College of Computer Science, Northeastern University, Boston, MA 02115, rraj@ccs.neu.edu. Supported by NSF CAREER award NSF CCR{9983901. 2 Department of Computer Science and Engineering, Arizona State University, Tempe, AZ 85287-5406, aricha@asu.edu. Supported in part by NSF CAREER Award CCR{9985284 and NSF Grant CCR{9900304. 3 Max-Planck-Institut f ur Informatik, Saarbr ucken, Germany. Supported in part by the IST Programme of the EU under contrac number IST-1999-14186 (ALCOM-FT) 4 Compaq Corporation, San Jose, CA, Gayathri.Vuppuluri@compaq.com. This work was done while the author was a graduate student at Arizona State University, supported in part by NSF CAREER Award CCR{9985284. 1

We consider the k-server problem [23] in a distributed setting.

Most of the work done in the area of static routing concentrates on minimizing the runtime of the whole schedule rather than minimizing the runtime of individual packets, using global parameters such as the congestion and dilation of a path collection. In this paper, we study the problem of minimizing the routing time of individual packets, using local parameters. In fact, we present the first (up to a log log factor) optimal, truly on-line routing protocols for the following problems: Packet switching: Assume that a fixed collection of paths is given. For every path p in this collection, let c p denote the maximum number of paths that share an edge with p, and let d p denote the length of p. Find a schedule (that does not require to know c p and d p ) such that the routing time of a packet following a path p merely depends on c p and d p . Virtual circuit switching: Assume that a fixed set of sessions is given. For every session i, packets are injected at a rate r i to f...

. We survey distributed data management problems including distributed paging, file allocation, and file migration. 1 Introduction Many modern information services know no national boundaries. The widespread use of the Internet and Internet-related applications such as the World Wide Web is growing fantastically on an annual basis. This survey deals with distributed data management problems. Such problems may arise as a memorymanagement problem for a globally addressed shared memory in a multiprocessor system as well as in a distributed network of processors where data files are kept in different sites and may be accessed for information retrieval by dispersed users and applications. In this context, a file may be a conventional single file, a system database, fragments of a database, or any combination of these. When a processor wishes to access a file it must send a request to a processor holding the file and the desired information is transmitted back. The communication cost in...

This paper deals with data management for parallel and distributed systems in which the computing nodes are connected by a relatively sparse network. We present the DIVA (Distributed Variables) library that provides fully transparent access to global variables, i.e., shared data objects, from the individual nodes in the network. The current implementations are based on mesh-connected massively parallel computers. The data management strategies implemented in the library use a non-standard approach based on a randomized but locality preserving embedding of “access trees ” into the physical network. The access tree strategy was previously analyzed only in a theoretical model using competitive analysis, where it was shown that the strategy produces minimal network congestion up to small factors. In this paper, the access tree strategy will be evaluated experimentally. We test several variations of this strategy on three different

The paper explores schemes for dynamic replication and migration of web objects in the context of an Internet hosting service. It describes a replica placement algorithm for deciding the location and number of replicas of an object as well as request distribution schemes for choosing among currently available replicas. We compare two classes of request distribution algorithms -- namely feedback and non-feedback based. Further, we compare dynamic replication to a static replication scheme. We have simulated the algorithms using synthetic workloads as well as a real trace from a hosting service. Measurement and analysis show that dynamic replication significantly reduces bandwidth consumption and latency, removes hot spots from the network and smooths out bursts in bandwidth demand while imposing only a low network traffic overhead. For example, on the trace, our algorithm reduces bandwidth consumption by as much as 52% while imposing a traffic overhead of only about ...