We investigate the simple class of greedy scheduling algorithms, that is, algorithms that always forward a packet if they can. Assuming that only one packet can be delivered over a link in a single step, and that the routes traversed by a set of packets are distance optimal ("shortest paths"), we prove that the time required to complete transmission of a packet in the set is bounded by its route length plus the number of other packets in the set. This bound holds for any greedy algorithm, even in the case of different starting times and different route lengths. The bound also generalizes, in the natural way, to the case in which w packets may cross a link simultaneously. Furthermore, the result holds in the asynchronous model, using the same proof technique. The generality of our result is demonstrated by some applications. We present a simple protocol, for which we derive a general bound on the throughput with any greedy scheduling. Another protocol for the dynamic case is ...

Traditionally, allocation of data in distributed database management systems has been determined by o�-line analysis and optimization. This technique works well for static database access patterns, but is often inadequate for frequently changing workloads. In this paper we address how to dynamically reallocate data for partionable distributed databases with changing access patterns. Rather than complicated and expensive optimization algorithms, a simple heuristic is presented and shown, via an implementation study, to improve system throughput by 30 � in a local area network based system. Based on arti�cial wide area network delays, we show that dynamic reallocation can improve system throughput by a factor of two and a half for wide area networks. We also show that individual site load must be taken into consideration when reallocating data, and provide a simple policy that incorporates load in the reallocation decision.

Stringent performance requirements in DB applications have led to the use of parallelism for database processing. To allow the database system to take advantage of the performance of parallel shared-nothing systems, the physical DB design must be appropriate for the DB structure and the workload. We develop decision algorithms that will select a good physical DB design both when the DB is first loaded into the system (static decision) and while the DB is being used by the workload (dynamic decision). Our decision algorithms take the database structure, workload, and system characteristics as inputs. The static (or initial) physical DB design decision algorithm involves: • selecting a partitioning attribute for each relation that determines how the relation is fragmented across the nodes (allowing for high I/O bandwidth); • selecting indexes on the relation attributes to allow faster accesses compared to sequential file scans; • selecting the attributes by which to cluster a relation in order to take advantage of the prefetching and caching involved in I/O access; • grouping of relations to allow DB operations (joins) on relation pairs to be executed locally

Abstract. A major cost in executing queries in a distributed database system is the data transfer cost incurred in transferring relations (fragments) accessed by a query from different sites to the site where the query is initiated. The objective of a data allocation algorithm is to determine an assignment of fragments at different sites so as to minimize the total data transfer cost incurred in executing a set of queries. This is equivalent to minimizing the average query execution time, which is of primary importance in a wide class of distributed conventional as well as multimedia database systems. The data allocation problem, however, is NP-complete, and thus requires fast heuristics to generate efficient solutions. Furthermore, the optimal allocation of database objects highly depends on the query execution strategy employed by a distributed database system, and the given query execution strategy usually assumes an allocation of the fragments. We develop a site-independent fragment dependency graph representation to model the dependencies among the fragments accessed by a query, and use it to formulate and tackle data allocation problems for distributed database systems based on query-site and move-small query execution strategies. We have designed and evaluated evolutionary algorithms for data allocation for distributed database systems.

This paper is concerned with fast, hot-potato routing, performed according to a predetermined schedule. At each time period each node selects an outgoing link, through which an incoming packet is sent. No buffers are used. We investigate first the problem of how to route a network-wide demand of packets, given the predetermined schedule. We show that certain versions of the problem have efficient solutions, while other versions are intractable. We then consider the problem of finding an optimal schedule given a network-wide demand of packets. We indicate that the problem is tractable for either a single source or single destination. However, for the multi-source multi-destination case we show that it is an NP-complete problem. We present an efficient heuristic for directed tree-networks, and adapt it to general topologies through a recursive scheme, for which an efficient performance bound is shown.

We consider the problem of transferring a set of files from their given locations in a fully connected network to their respective destinations in minimum time. The network has two unidirectional links, one in each direction, between every pair of nodes. We assume that it takes unit time to transfer a file across a link and no link is used by more than one file at any time. There is no restriction on storage capacity at the nodes. The objective of the File Redistribution Scheduling problem is to find routes for the files and a schedule for the use of the links along the routes so as to complete the transfer of all the files in minimum time. This problem has been shown to be NP-hard even with the restriction that each file must have at most one hop in its route. In this paper, we present an efficient polynomial time algorithm that finds an approximate solution to the problem. In this approximate solution, each file has at most one hop in its route and all the files can be tr...

A major cost in executing queries in a distributed database system is the data transfer cost incurred in transferring multiple database objects (fragments) accessed by a query from different sites to the site where the query is initiated. The objective of a data allocation algorithm is to locate the fragments at different sites so as to minimize the total data transfer cost incurred in executing a set of queries. We develop a site-independent fragment dependency graph representation to model the dependencies among the fragments accessed by a query, and use it to formulate and solve data allocation problems for distributed database systems based on (query-site and move-small) query execution strategies. We show that an optimal solution can be achieved when the query-site query execution strategy is employed, and for the move-small query execution strategy we performed experimental evaluation about the effectiveness of a hillclimbing heuristic algorithm in achieving a near-optimal soluti...

An application processing center consists of a set of well-defined, well-designed and well-tested applications that are dynamically executed over a period of time. We assume that there is a set of candidate distributed database designs each of which is optimal for some applications. The random execution of applications on a distributed database design is modeled as a discrete Markov process, and the problem of selecting the candidate design for each execution of an application is solved by using sequential Markovian decision process analysis to generate an optimal redesign policy vector. The scope of the methodology developed in this paper is applicable to environments similar to application processing centers. The viability of this methodology is illustrated by means of a case study conducted at Georgia Tech. 1 Introduction Relational database system technology has been under development for more than two decades. There are many efficient commercial relational database systems in exi...

this paper, we address the problem of stepwise