Parallel computers have been successfully deployed in many scientific and numerical application areas, although their use in non-numerical and database applications has been scarce. In this report, we first survey the architectural advancements beginning to make general-purpose parallel computing cost-effective, the requirements for non-numerical (or symbolic) applications, and the previous attempts to develop parallel databases. The central theme of the Bulk Synchronous Parallel model is to provide a high level abstraction of parallel computing hardware whilst providing a realisation of a parallel programming model that enables architecture independent programs to deliver scalable performance on diverse hardware platforms. Therefore, the primary objective of this report is to investigate the feasibility of developing a portable, scalable, parallel object database, based on the Bulk Synchronous Parallel model of computation. In particular, we devise a way of providing high-level abstra...

Parallel computing is becoming increasingly accessible through advancement in microprocessors and networking technologies. It is found that the performances of the clusters do not match with their promises, although they are built on the most advanced components. Much effort has been devoted to address the software overhead problem in the past, which is known as the major hindrance in achieving high performance. This thesis shows that having a low-latency communication system does not guarantee high performance, as there are other communication issues that have not been addressed by the use of low-latency communication, such as contention, communication patterns and scheduling of communication events. The development of an efficient parallel application depends upon a realistic prediction of application behavior and the ability to explain the performance characteristics of an application on a parallel system; this requires in-depth understanding of both the application and the architecture characteristics. This dissertation proposes the use of a realistic communication model to guide the performance understanding and the algorithm design processes, which are the keys to

) Armin Baumker, Wolfgang Dittrich and Friedhelm Meyer auf der Heide Department of Mathematics and Computer Science and Heinz Nixdorf Institute, University of Paderborn 33095 Paderborn, Germany Abstract In this paper we design and analyse parallel algorithms with the goal to get exact bounds on their speed-ups on real machines. For this purpose we define an extension of Valiant's BSP model, BSP*, that rewards blockwise communication, and uses Valiant's notion of c-optimality. Intuitively a c-optimal parallel algorithm for p processors achieves speed-up close to p=c. We consider the Multisearch problem: Assume a strip in 2D to be partitioned into m segments. Given n query points in the strip, the task is to locate, for each query, its segment. For m n we present a deterministic BSP* algorithm that is 1-optimal, if n =\Omega (p log 2 p). For m ? n, we present a randomized BSP* algorithm that is (1 + ffi)-optimal for arbitrary ffi ? 0, m 2 p and n =\Omega (p log 2 p). Both r...

and there currently exist a wide selection of parallel programming languages and environments. This thesis presents and examines the Hierarchical Skeleton Model (HSM), a model of parallel programming that combines ease of use, portability and flexibility.

. Searching is one of the most important algorithmic problems, used as a subroutine in many applications. Accordingly, designing search algorithms is in the center of research on data structures since decades. In this paper we aim to survey recent developments in designing parallel search algorithms where parallel machines are used to answer many search queries in parallel, so called multisearch algorithms. We briefly describe the current state of multisearch algorithms based on hashing and binary search, as they are developed for abstract parallel models like the PRAM. The main part of the paper describes deterministic and randomized multisearch algorithms that are very communication efficient. As a computation and cost model we employ Valiant's BSP model and its variant BSP* due to Baumker et al. 1 Introduction One of the most important algorithmic problems is that of searching in a given set of objects where each object can be identified by a search key. In order to illustrate the ...

Bulk Synchronous Parallelism (BSP) as a parallel model enables accurate costs of parallel programs to be predicted from the program structure and two architectural parameters, g, the permeability of the network, and l, the time required for barrier synchronisation. Networks such as ATM already play a role in parallel computers built as networks of workstations, and may become the standard mechanism for interconnecting processors at all scales. We present an analytic model for determining the BSP parameters of such architectures. Although the model is simple, there is substantial agreement with measured results where these are known. This represents the first time that these architectural parameters have been determined other than by benchmarking, and suggests that the approach may be serviceable for other wormhole routed networks. Keywords: parallel computing, interconnection network, performance modelling, total exchange, bulk synchronous parallelism, latency, throughput. 1 Introd...

In this paper we design and analyse parallel algorithms with the goal to get exact bounds on their speed-ups on real machines. For this purpose we define an extension of Valiant's BSP model, BSP*, that rewards blockwise communication, and use Valiant's notion of 1-optimality. Intuitively, a 1-optimal parallel algorithm for p processors achieves speed-up close to p. We consider the Multisearch Problem: Assume a strip in 2D to be partitioned into m segments. Given n query points in the strip, the task is to locate, for each query, its segment. For m n p we present a deterministic BSP* algorithm that is 1-optimal, if n p log 2 n. For m ? n p, we present a randomized BSP* algorithm that is 1-optimal with high probability, if m 2 p and n p log 3 n. Both results hold for a wide range of BSP* parameters where the range becomes larger with growing input size n. We further report on implementation work. Previous parallel algorithms for Multisearch were far away from being 1-op...

A computer framework feasible for developing parallel systems according to the Bulk Synchronous Parallel (BSP) computing model is described: Switched Interconnection of Parallel Processors (Swipp). Demanding applications can be described as directed graphs where the interdependent subtasks constitute the nodes. The tasks are predistributed by a system master, Computer Executive Engine (CEE), to a set of heterogeneous computing nodes. Each computing node has a preprogrammed secondary control processor attached for performing communication and runtime tasks, thus allowing primary processors of various kinds and programming styles. Synchronization of a bulk of subactivities is done in locksteps by the CEE. Basic features are modelled by the Ptolemy framework and prototype modules are being implemented. I. INTRODUCTION The performance of sequential single CPU processors has been steadily increasing over the last decades due to circuit technology improvements. The von Neumann model of comp...

MACHINES (COMPUTATIONAL MODELS) ............................. 4 2.2 GENERAL PROPERTIES OF PARALLEL COMPUTATIONAL MODELS.................................................... 4 2.3 THE PRAM MODEL AND ITS LINK TO THE BSP MODEL .......................................................................... 5 2.3 BRIDGING MODELS IN GENERAL................................................................................................................. 6 3. VALIANT'S BULK SYNCHRONOUS PARALLEL (BSP) MODEL ................................................................. 7 3.1 VALIANT'S ORIGINAL BSP MODEL.............................................................................................................. 7 3.2 DESIGN CHOICES IN THE OXFORD BSP MODEL (AND IN BSPLAB) ...................................................... 7 One synchronization call ends a superstep ............................................................................................................ 8 Interference-free sup...

A major challenge for parallel computing is the development of a standardized combination of portable and efficient parallel programming. An interesting approach towards this major goal is the research with offspring in Leslie Valiant's Bulk Synchronous Parallel Model (BSP). The BSP model is a theoretical framework outlining how parallel computations can be organized in a way that bridges the gap between the needs of the programmers and the technological possibilities of the computer designers. BSPlab is an environment that makes it possible to develop, debug and measure BSP programs on a variety of parallel computer architectures. The typical use of BSPlab is experiments to learn about the interplay between hardware and software in the search for efficient and portable parallel programming. The project is an offspring from studies on computational models and an interest in using results from theoretical computer science in practical computer engineering. The paper starts by...