Browsing by Subject "Parallel processing (Electronic computers)."

Now showing 1 - 8 of 8

Open Access
Adaptive source routing and route generation for multicomputers
(1995) Aydoğan, Yücel
Scalable multicomputers are based upon interconnection networks that typically provide multiple communication routes between any given pair of processor nodes. In such networks, the selection of the routes is an important problem because of its impact on the communication performance. We propose the adaptive source routing (ASR) scheme which combines adaptive routing and source routing into one which has the advantages of both schemes. In ASR, the degree of adaptivity of each packet is determined at the source processor. Every packet can be routed in a fully adaptive or partially adaptive or nonadaptive manner, all within the same network at the same time. The ASR scheme permits any network topology to be used provided that deadlock constraints are satisfied. We evaluate and compare performance of the adaptive source routing and non-adaptive randomized routing by simulations. Also we propose an algorithm to generate adaptive routes for all pairs of processors in any multistage interconnection network. Adaptive routes are stored in a route table in each processor’s memory and provide high bandwidth and reliable interprocessor communication. We evaluate the performance of the algorithm on IBM SP2 networks in terms of obtained bandwidth, time to fill in the route tables, and efficiency exploited by the parallel execution of the algorithm.
Open Access
Distributed scheduling
(1999) Toptal, Ayşegül
Distributed Scheduling (DS) is a new paradigm that enables the local decisionmakers make their own schedules by considering local objectives and constraints within the boundaries and the overall objective of the whole system. Local schedules from different parts of the system are then combined together to form a final schedule. Since each local decision-maker acts independently from each other, the communication system in a distributed architecture should be carefully designed to achieve better overall system performance. These systems are preferred over the traditional systems due to the ability to update the schedule, flexibility, reactivity and shorter lead times. In this thesis, we review the existing work on DS and propose a new classification framework. We also develop a number of bidding based DS algorithms. These algorithms are tested under various manufacturing environments.
Open Access
Implementation of the backpropagation algorithm on iPSC/2 hypercube multicomputer system
(1990) Ercoşkun, Deniz
Backpropagation is a supervised learning procedure for a class of artificial neural networks. It has recently been widely used in training such neural networks to perform relatively nontrivial tasks like text-to-speech conversion or autonomous land vehicle control. However, the slow rate of convergence of the basic backpropagation algorithm has limited its application to rather small networks since the computational requirements grow significantly as the network size grows. This thesis work presents a parallel implementation of the backpropagation learning algorithm on a hypercube multicomputer system. The main motivation for this implementation is the construction of a parallel training and simulation utility for such networks, so that larger neural network applications can be experimented with.
Open Access
Models and algorithms for parallel text retrieval
(2006) Cambazoğlu, Berkant Barla
In the last decade, search engines became an integral part of our lives. The current state-of-the-art in search engine technology relies on parallel text retrieval. Basically, a parallel text retrieval system is composed of three components: a crawler, an indexer, and a query processor. The crawler component aims to locate, fetch, and store the Web pages in a local document repository. The indexer component converts the stored, unstructured text into a queryable form, most often an inverted index. Finally, the query processing component performs the search over the indexed content. In this thesis, we present models and algorithms for efficient Web crawling and query processing. First, for parallel Web crawling, we propose a hybrid model that aims to minimize the communication overhead among the processors while balancing the number of page download requests and storage loads of processors. Second, we propose models for documentand term-based inverted index partitioning. In the document-based partitioning model, the number of disk accesses incurred during query processing is minimized while the posting storage is balanced. In the term-based partitioning model, the total amount of communication is minimized while, again, the posting storage is balanced. Finally, we develop and evaluate a large number of algorithms for query processing in ranking-based text retrieval systems. We test the proposed algorithms over our experimental parallel text retrieval system, Skynet, currently running on a 48-node PC cluster. In the thesis, we also discuss the design and implementation details of another, somewhat untraditional, grid-enabled search engine, SE4SEE. Among our practical work, we present the Harbinger text classification system, used in SE4SEE for Web page classification, and the K-PaToH hypergraph partitioning toolkit, to be used in the proposed models.
Open Access
Parallel text retrieval on PC clusters
(2003) Çatal, Aytül
The inverted index partitioning problem is investigated for parallel text retrieval systems. The objective is to perform efficient query processing on an inverted index distributed across a PC cluster. Alternative strategies are considered and evaluated for inverted index partitioning, where index entries are distributed according to their document-ids or term-ids. The performance of both partitioning schemes depend on the total number of disk accesses and the total volume of communication in the system. In document-id partitioning, the total volume of communication is naturally minimum, whereas the total number of disk accesses may be larger compared to term-id partitioning. On the other hand, in term-id partitioning the total number of disk accesses is already equivalent to the lower bound achieved by the sequential algorithm, albeit the total communication volume may be quite large. The studies done so far perform these partitioning schemes in a round-robin fashion and compare the performance of them by simulation. In this work, a parallel text retrieval system is designed and implemented on a PC cluster. We adopted hypergraph-theoretical partitioning models and carried out performance comparison of round-robin and hypergraph-theoretical partitioning schemes on our parallel text retrieval system. We also designed and implemented a query interface and a user interface of our system.
Open Access
Parallel text retrieval on temporally versioned document collections
(2008) Gür, Özlem
In recent years, as the access to the Internet is getting easier and cheaper, the amount and the rate of change of the online data presented to the Internet users are increasing at an astonishing rate. This ever-changing nature of the Internet causes an ever-decaying and replenishing information collection where newly presented data generally replaces old and sometimes valuable data. There are many recent studies aiming to preserve this valuable temporal data and size and number of temporal Web data collections are increasing. We believe that soon, information retrieval systems responding to time-range queries in a reasonable amount of time will emerge as a means of accessing vast temporal Web data collections. Due to tremendous size of temporal data and excessive number of query submissions per unit time, temporal information retrieval systems will have to utilize parallelism as much as possible. In parallel systems, in order to index collections using inverted indices, a strategy on distribution of the inverted indices has to be followed. In this study, the feasibility of time-based partitioned versus term-based partitioned temporalweb inverted-indices is analyzed and a novel parallel text retrieval system for answering temporal web queries is implemented considering the number of queries processed in unit time. Moreover, we investigate the performance of skip-list based and randomized-select based ranking schemes on time-based and termbased partitioned inverted indexes. Finally, we compare time-balanced and sizebalanced time-based partitioning schemes. The experimental results at small to medium number of processors reveal that for medium to long length queries time-based partitioning works better.
Open Access
Performance comparison of query evaluation techniques in parallel text retrieval systems
(2008) Tokuç, A. Aylin
Today’s state-of-the-art search engines utilize the inverted index data structure for fast text retrieval on large document collections. To parallelize the retrieval process, the inverted index should be distributed among multiple index servers. Generally the distribution of the inverted index is done in either a term-based or a document-based fashion. The performances of both schemes depend on the total number of disk accesses and the total volume of communication in the system. The classical approach for both distributions is to use the Central Broker Query Evaluation Scheme (CB) for parallel text retrieval. It is known that in this approach the central broker is heavily loaded and becomes a bottleneck. Recently, an alternative query evaluation technique, named Pipelined Query Evaluation Scheme (PPL), has been proposed to alleviate this problem by performing the merge operation on the index servers. In this study, we analyze the scalability and relative performances of the CB and PPL under various query loads to report the benefits and drawbacks of each method.
Open Access
Spatial subdivision for parallel ray casting/tracing
(1995) İşler, Veysi
Ray casting/tracing has been extensively studied for a long time, since it is an elegant way of producing realistic images. However, it is a computationally intensive algorithm. In this study, a taxonomy of parallel ray casting/tracing algorithms is presented cind the primary parallel ray casting/tracing systems are discussed and criticized. This work mainly focuses on the utilization of spatial subdivision technique for ray casting/tracing on a distributed-memory MIMD parallel computer. In this research, the reason for the use of parallel computers is not only the processing power but also the large memory space provided by them. The spatial subdivision technique has been adapted to parallel ray casting/tracing to decompose a three-dimensional complex scene that may not fit into the local memory of a single processor. The decomposition method achieves an even distribution of scene objects while allowing to exploit graphical coherence. Additionally, the decomposition method produces three-dimensional volumes which are mapped inexpensively to the processors so that the objects within adjacent volumes are stored in the local memories of close processors to decrease interprocessor communication cost. Then, the developed decomposition and mapping methods have been parallelized efficiently to reduce the preprocessing overhead. Finally, a splitting plane concept (called “jaggy splitting plane”) has been proposed to accomplish full utilization of the memory space of processors. Jaggy splitting plane avoids the shared objects which are the major sources of inefficient utilization of both memory and processing power. The proposed parallel algorithms have been implemented on the Intel iPSC/2 hypercube multicomputer (distributed-memory MIMD).