Skip to main content

DongSeung Kim

Followers

6

Following

1

Public Views

Aswani Kumar Cherukuri

VIT University

Armando Marques-Guedes

UNL - New University of Lisbon

Kati (Katalin) Prajda

University of Vienna

Graduate Center of the City University of New York

Annajiat Alim Rasel

BRAC University

Journal of Computer Science IJCSIS

Ain Shams University

PES University

Leading University

Mohd Atif Kaleem

Aligarh Muslim University

Interests

Uploads

Papers by DongSeung Kim

Quick Community Detection of Big Graph Data Using Modified Louvain Algorithm

Louvain algorithm is known to be fast to deliver good results in detecting communities of big gra... more Louvain algorithm is known to be fast to deliver good results in detecting communities of big graphs. This paper further speeds up Louvain algorithm by restricting the internal search rules and early pruning the non-promising candidates. Experimental results of the modified algorithm on various sized data show outstanding speedups of up to 40.5 and 4.7 times for weighted graphs and unweighted graphs, respectively.

Weather-Aware Long-Range Traffic Forecast Using Multi-Module Deep Neural Network

Applied Sciences

This study proposes a novel multi-module deep neural network framework which aims at improving in... more This study proposes a novel multi-module deep neural network framework which aims at improving intelligent long-term traffic forecasting. Following our previous system, the internal architecture of the new system adds deep learning modules that enable data separation during computation. Thus, prediction becomes more accurate in many sections of the road network and gives dependable results even under possible changes in weather conditions during driving. The performance of the framework is then evaluated for different cases, which include all plausible cases of driving, i.e., regular days, holidays, and days involving severe weather conditions. Compared with other traffic predicting systems that employ the convolutional neural networks, k-nearest neighbor algorithm, and the time series model, it is concluded that the system proposed herein achieves better performance and helps drivers schedule their trips well in advance.

Generalized Processor-Memory Bandwidth Analysis For Fully Shared-Memory Multiprocessors

Twenty-Second Asilomar Conference on Signals, Systems and Computers

Effective bandwidths of crossbar and multiple bus interconnections applied for shared-memory mult... more Effective bandwidths of crossbar and multiple bus interconnections applied for shared-memory multiprocessors are analyzed. The analysis uses general access patterns among processors and memories. Bandwidth functions of the interconnections are mathematically proven increasing with respect to the number of buses and memory access rates. The minimum bandwidth is obtained when uniform access is assumed, and the maximum when memory references are concentrated to one memory module for each processor. Even if some of the access probabilities are given fixed, the minimum and maximum are similarly found when the remaining access probabilities are equal and concentrated, respectively.

A Dynamic Hashing Based Load Balancing for a Scalable Wireless Internet Proxy Server Cluster

The KIPS Transactions:PartA

Load Balanced Block Lanczos Algorithm over GF(2) for Factorization of Large Keys

Lecture Notes in Computer Science, 2006

Researchers use NFS (Number Field Sieve) method with Lanczos algorithm to analyze big-sized RSA k... more Researchers use NFS (Number Field Sieve) method with Lanczos algorithm to analyze big-sized RSA keys. NFS method includes the integer factorization process and nullspace computation of huge sparse matrices. Parallel processing is indispensible since sequential computation requires weeks (even months) of CPU time with supercomputers even for 150-digit RSA keys. This paper presents details of improved block Lanczos algorithm based on previous implementation . It includes a new load balancing scheme by partitioning the matrix such that the numbers of nonzero components in the submatrices become equal. Experimentally, a speedup up to 6 and the maximum of efficiency of 0.74 have been achieved using an 8node cluster with Myrinet interconnection.

DTN: A New Partitionable Torus Topology

Fast Broadcast by Message Segmentation

In this paper we propose a fast broadcast scheme called segmented broadcast, and its performance ... more In this paper we propose a fast broadcast scheme called segmented broadcast, and its performance i s a n a l y z e d and compared to known schemes. Experiments are p erformed to verify the validity of it on IBM SP2 and Cray T3E under MPI{1 environment. The results turn out that the speedup enhancement for long messages is up to two-fold or more, as the number of processors grows. It can also be applied to broadcast under MPI{2 or similar message passing environments without any modi cation. The concept is being extended to apply to other communication functions such as scatter, gather, and scan.

External sort on a cluster of PCs

Parallel Pattern ClusterIng on a Multiprocessor with Orthogonally Shared Memory

Distribution Based on Client Cache Validation

International Conference on Internet Computing, 2002

Orthogonal multiprocessor sharing memory with an enhanced mesh for integrated image understanding

CVGIP: Image Understanding, 1991

Abstract This paper proposes a new parallel architecture, which has the potential to support low-... more

High-speed parallel external sorting of data with arbitrary distribution

International Journal of High Performance Computing and Networking, 2004

Many parallel sorting algorithms of (external) disk data have been reported such as NOWsort, SPso... more Many parallel sorting algorithms of (external) disk data have been reported such as NOWsort, SPsort, and hill sort, etc. They all reduce the execution time compared to some known sequential sort; however, they differ in terms of the speed, throughput, and cost-effectiveness. Mostly they deal with data that are uniformly distributed in their value range. If we divide and redistribute data to processors using the fixed and equal division of the key range, all processors will have about equal number of keys to sort and store. But if irregularly distributed data are given, the performance will suffer severely since the partitioning would no longer produce balanced load among processors. Few research results have been yet reported for parallel external sort of data with arbitrary distribution. In this paper, we develop two distribution-insensitive scalable parallel external sorting algorithms that use sampling technique and histogram counts to achieve even distribution of data, which eventually contribute to achieve superb performance. Experimental results on a cluster of 16 Linux workstations show up to threefold enhancement of the performance compared to previous NOW-sort for sorting 16GB integer keys.

Distribution-Insensitive Parallel External Sorting on PC Clusters

Lecture Notes in Computer Science, 2003

There have been many parallel external sorting algorithms reported such as NOW-Sort, SPsort, and ... more There have been many parallel external sorting algorithms reported such as NOW-Sort, SPsort, and hill sort, etc. They are for sorting large-scale data stored in the disk, but they differ in the speed, throughput, and cost-effectiveness. Mostly they deal with data that are uniformly distributed in their value range. Few research results have been yet reported for parallel external sort

Parallelizing Merge Sort onto Distributed Memory Parallel Computers

Lecture Notes in Computer Science, 2002

Merge sort is useful in sorting a great number of data progressively, especially when they can be... more Merge sort is useful in sorting a great number of data progressively, especially when they can be partitioned and easily collected to a few processors. Merge sort can be parallelized, however, conventional algorithms using distributed memory computers have poor performance due to the successive reduction of the number of participating processors by a half, up to one in the last merging stage. This paper presents load-balanced parallel merge sort where all processors do the merging throughout the computation. Data are evenly distributed to all processors, and every processor is forced to work in all merging phases. An analysis shows the upper bound of the speedup of the merge time as (P − 1)/ log P where P is the number of processors. We have reached a speedup of 8.2 (upper bound is 10.5) on 32-processor Cray T3E in sorting of 4M 32-bit integers.

Load-balanced parallel merge sort on distributed memory parallel computers

Proceedings 16th International Parallel and Distributed Processing Symposium, 2002

Sort can be speeded up on parallel computers by dividing and computing data individually in paral... more Sort can be speeded up on parallel computers by dividing and computing data individually in parallel. Merge sort can be parallelized, however, the conventional algorithm implemented on distributed memory computers has poor performance due to the successive reduction of the number of active (non-idling) processors by a half, up to one in the last merging stage. This paper presents load-balanced parallel merge sort algorithm where all processors participate in merging throughout the computation. Data are evenly distributed to all processors, and every processor is forced to work in merging phase. Significant enhancement of the performance has been achieved. Our analysis shows the upper bound of the speedup of the merge time as (P ;1)= logP. We have had a speedup of 9.6 (upper bound is 10.5) on 32-processor Cray T3E in sorting of 4M 32-bit integers. The same idea can be applied to parallellize other sorting algorithms.

Load Balanced Parallel Prime Number Generator with Sieve of Eratosthenes on Cluster Computers

7th IEEE International Conference on Computer and Information Technology (CIT 2007), 2007

Algorithms of finding prime numbers up to some integer N by Sieve of Eratosthenes are simple and ... more Algorithms of finding prime numbers up to some integer N by Sieve of Eratosthenes are simple and fast. However, even with the time complexity no greater than O(N ln ln N), it may take weeks or even years of CPU time if N is large like over 15 digits. No known shortcut was reported yet. We develop efficient parallel algorithms to balance the workload of each computer, and to extend main memory with disk storage to accommodate Giga-bytes of data. Our experiments show that a complete set of up to 14-digit prime numbers can be found in a week of computing time (CPU time) using our 8 1.6GHz Pentium-4 PCs with Linux and MPI library. Also, by sieve of Eratosthenes, we think it is very unlikely that we can compute all primes up to 20 digits using fastest computers in the world.

Demonstration of Location Information Based Network

2008 International Conference on Information Processing in Sensor Networks (ipsn 2008), 2008

In ubiquitous network, node location information is the most essential data for efficiently manag... more In ubiquitous network, node location information is the most essential data for efficiently managing increased mobility and supporting location based services. Therefore, we propose a next generation location information based network, which uses address containing location information. Finally, its functions, effectiveness, and feasibility will be verified through this demonstration.

Hierarchical cluster for scalable web servers

Cluster architecture is widely used for the availability and scalability of web servers in these ... more Cluster architecture is widely used for the availability and scalability of web servers in these days. To increase the performance we have developed a hierarchical architecture for web servers that can accommodate wide range of connection requests by gradually expanding the nodes in proper levels in the hierarchy. Especially, multiple dispatchers can distribute the incoming client requests to various backend servers in the cluster. Experimental results with two-level cluster with dynamic load balancing are included to verify the idea.

Performance analysis and experiments of sorts on a parallel computer with parallel computation models

A Method for Optimal Bandwidth Utilization in IEEE 802.11 WLAN Networks

2008 International Conference on Information Networking, 2008

This paper proposes a load sharing scheme to maximize network bandwidth utilization in IEEE 802.1... more This paper proposes a load sharing scheme to maximize network bandwidth utilization in IEEE 802.11 WLAN networks using the SSED (Service Set Identifier) hiding. For this purpose, the proposed scheme keeps checking the available bandwidths of a group of wireless routers; selects the most bandwidth-optimal one; makes it visible to clients. Such that, whenever a client connects to a wireless

Quick Community Detection of Big Graph Data Using Modified Louvain Algorithm

Louvain algorithm is known to be fast to deliver good results in detecting communities of big gra... more Louvain algorithm is known to be fast to deliver good results in detecting communities of big graphs. This paper further speeds up Louvain algorithm by restricting the internal search rules and early pruning the non-promising candidates. Experimental results of the modified algorithm on various sized data show outstanding speedups of up to 40.5 and 4.7 times for weighted graphs and unweighted graphs, respectively.

Weather-Aware Long-Range Traffic Forecast Using Multi-Module Deep Neural Network

Applied Sciences

This study proposes a novel multi-module deep neural network framework which aims at improving in... more This study proposes a novel multi-module deep neural network framework which aims at improving intelligent long-term traffic forecasting. Following our previous system, the internal architecture of the new system adds deep learning modules that enable data separation during computation. Thus, prediction becomes more accurate in many sections of the road network and gives dependable results even under possible changes in weather conditions during driving. The performance of the framework is then evaluated for different cases, which include all plausible cases of driving, i.e., regular days, holidays, and days involving severe weather conditions. Compared with other traffic predicting systems that employ the convolutional neural networks, k-nearest neighbor algorithm, and the time series model, it is concluded that the system proposed herein achieves better performance and helps drivers schedule their trips well in advance.

Generalized Processor-Memory Bandwidth Analysis For Fully Shared-Memory Multiprocessors

Twenty-Second Asilomar Conference on Signals, Systems and Computers

Effective bandwidths of crossbar and multiple bus interconnections applied for shared-memory mult... more Effective bandwidths of crossbar and multiple bus interconnections applied for shared-memory multiprocessors are analyzed. The analysis uses general access patterns among processors and memories. Bandwidth functions of the interconnections are mathematically proven increasing with respect to the number of buses and memory access rates. The minimum bandwidth is obtained when uniform access is assumed, and the maximum when memory references are concentrated to one memory module for each processor. Even if some of the access probabilities are given fixed, the minimum and maximum are similarly found when the remaining access probabilities are equal and concentrated, respectively.

A Dynamic Hashing Based Load Balancing for a Scalable Wireless Internet Proxy Server Cluster

The KIPS Transactions:PartA

Load Balanced Block Lanczos Algorithm over GF(2) for Factorization of Large Keys

Lecture Notes in Computer Science, 2006

Researchers use NFS (Number Field Sieve) method with Lanczos algorithm to analyze big-sized RSA k... more Researchers use NFS (Number Field Sieve) method with Lanczos algorithm to analyze big-sized RSA keys. NFS method includes the integer factorization process and nullspace computation of huge sparse matrices. Parallel processing is indispensible since sequential computation requires weeks (even months) of CPU time with supercomputers even for 150-digit RSA keys. This paper presents details of improved block Lanczos algorithm based on previous implementation . It includes a new load balancing scheme by partitioning the matrix such that the numbers of nonzero components in the submatrices become equal. Experimentally, a speedup up to 6 and the maximum of efficiency of 0.74 have been achieved using an 8node cluster with Myrinet interconnection.

DTN: A New Partitionable Torus Topology

Fast Broadcast by Message Segmentation

In this paper we propose a fast broadcast scheme called segmented broadcast, and its performance ... more In this paper we propose a fast broadcast scheme called segmented broadcast, and its performance i s a n a l y z e d and compared to known schemes. Experiments are p erformed to verify the validity of it on IBM SP2 and Cray T3E under MPI{1 environment. The results turn out that the speedup enhancement for long messages is up to two-fold or more, as the number of processors grows. It can also be applied to broadcast under MPI{2 or similar message passing environments without any modi cation. The concept is being extended to apply to other communication functions such as scatter, gather, and scan.

External sort on a cluster of PCs

Parallel Pattern ClusterIng on a Multiprocessor with Orthogonally Shared Memory

Distribution Based on Client Cache Validation

International Conference on Internet Computing, 2002

Orthogonal multiprocessor sharing memory with an enhanced mesh for integrated image understanding

CVGIP: Image Understanding, 1991

Abstract This paper proposes a new parallel architecture, which has the potential to support low-... more

High-speed parallel external sorting of data with arbitrary distribution

International Journal of High Performance Computing and Networking, 2004

Many parallel sorting algorithms of (external) disk data have been reported such as NOWsort, SPso... more Many parallel sorting algorithms of (external) disk data have been reported such as NOWsort, SPsort, and hill sort, etc. They all reduce the execution time compared to some known sequential sort; however, they differ in terms of the speed, throughput, and cost-effectiveness. Mostly they deal with data that are uniformly distributed in their value range. If we divide and redistribute data to processors using the fixed and equal division of the key range, all processors will have about equal number of keys to sort and store. But if irregularly distributed data are given, the performance will suffer severely since the partitioning would no longer produce balanced load among processors. Few research results have been yet reported for parallel external sort of data with arbitrary distribution. In this paper, we develop two distribution-insensitive scalable parallel external sorting algorithms that use sampling technique and histogram counts to achieve even distribution of data, which eventually contribute to achieve superb performance. Experimental results on a cluster of 16 Linux workstations show up to threefold enhancement of the performance compared to previous NOW-sort for sorting 16GB integer keys.

Distribution-Insensitive Parallel External Sorting on PC Clusters

Lecture Notes in Computer Science, 2003

There have been many parallel external sorting algorithms reported such as NOW-Sort, SPsort, and ... more There have been many parallel external sorting algorithms reported such as NOW-Sort, SPsort, and hill sort, etc. They are for sorting large-scale data stored in the disk, but they differ in the speed, throughput, and cost-effectiveness. Mostly they deal with data that are uniformly distributed in their value range. Few research results have been yet reported for parallel external sort

Parallelizing Merge Sort onto Distributed Memory Parallel Computers

Lecture Notes in Computer Science, 2002

Merge sort is useful in sorting a great number of data progressively, especially when they can be... more Merge sort is useful in sorting a great number of data progressively, especially when they can be partitioned and easily collected to a few processors. Merge sort can be parallelized, however, conventional algorithms using distributed memory computers have poor performance due to the successive reduction of the number of participating processors by a half, up to one in the last merging stage. This paper presents load-balanced parallel merge sort where all processors do the merging throughout the computation. Data are evenly distributed to all processors, and every processor is forced to work in all merging phases. An analysis shows the upper bound of the speedup of the merge time as (P − 1)/ log P where P is the number of processors. We have reached a speedup of 8.2 (upper bound is 10.5) on 32-processor Cray T3E in sorting of 4M 32-bit integers.

Load-balanced parallel merge sort on distributed memory parallel computers

Proceedings 16th International Parallel and Distributed Processing Symposium, 2002

Sort can be speeded up on parallel computers by dividing and computing data individually in paral... more Sort can be speeded up on parallel computers by dividing and computing data individually in parallel. Merge sort can be parallelized, however, the conventional algorithm implemented on distributed memory computers has poor performance due to the successive reduction of the number of active (non-idling) processors by a half, up to one in the last merging stage. This paper presents load-balanced parallel merge sort algorithm where all processors participate in merging throughout the computation. Data are evenly distributed to all processors, and every processor is forced to work in merging phase. Significant enhancement of the performance has been achieved. Our analysis shows the upper bound of the speedup of the merge time as (P ;1)= logP. We have had a speedup of 9.6 (upper bound is 10.5) on 32-processor Cray T3E in sorting of 4M 32-bit integers. The same idea can be applied to parallellize other sorting algorithms.

Load Balanced Parallel Prime Number Generator with Sieve of Eratosthenes on Cluster Computers

7th IEEE International Conference on Computer and Information Technology (CIT 2007), 2007

Algorithms of finding prime numbers up to some integer N by Sieve of Eratosthenes are simple and ... more Algorithms of finding prime numbers up to some integer N by Sieve of Eratosthenes are simple and fast. However, even with the time complexity no greater than O(N ln ln N), it may take weeks or even years of CPU time if N is large like over 15 digits. No known shortcut was reported yet. We develop efficient parallel algorithms to balance the workload of each computer, and to extend main memory with disk storage to accommodate Giga-bytes of data. Our experiments show that a complete set of up to 14-digit prime numbers can be found in a week of computing time (CPU time) using our 8 1.6GHz Pentium-4 PCs with Linux and MPI library. Also, by sieve of Eratosthenes, we think it is very unlikely that we can compute all primes up to 20 digits using fastest computers in the world.

Demonstration of Location Information Based Network

2008 International Conference on Information Processing in Sensor Networks (ipsn 2008), 2008

In ubiquitous network, node location information is the most essential data for efficiently manag... more In ubiquitous network, node location information is the most essential data for efficiently managing increased mobility and supporting location based services. Therefore, we propose a next generation location information based network, which uses address containing location information. Finally, its functions, effectiveness, and feasibility will be verified through this demonstration.

Hierarchical cluster for scalable web servers

Cluster architecture is widely used for the availability and scalability of web servers in these ... more Cluster architecture is widely used for the availability and scalability of web servers in these days. To increase the performance we have developed a hierarchical architecture for web servers that can accommodate wide range of connection requests by gradually expanding the nodes in proper levels in the hierarchy. Especially, multiple dispatchers can distribute the incoming client requests to various backend servers in the cluster. Experimental results with two-level cluster with dynamic load balancing are included to verify the idea.

Performance analysis and experiments of sorts on a parallel computer with parallel computation models

A Method for Optimal Bandwidth Utilization in IEEE 802.11 WLAN Networks

2008 International Conference on Information Networking, 2008

This paper proposes a load sharing scheme to maximize network bandwidth utilization in IEEE 802.1... more This paper proposes a load sharing scheme to maximize network bandwidth utilization in IEEE 802.11 WLAN networks using the SSED (Service Set Identifier) hiding. For this purpose, the proposed scheme keeps checking the available bandwidths of a group of wireless routers; selects the most bandwidth-optimal one; makes it visible to clients. Such that, whenever a client connects to a wireless