High performance data clustering: a comparative analysis of performance for GPU, RASC, MPI, and OpenMP implementations

Abstract

Compared to Beowulf clusters and shared-memory machines, GPU and FPGA are emerging alternative architectures that provide massive parallelism and great computational capabilities. These architectures can be utilized to run compute-intensive algorithms to analyze ever-enlarging datasets and provide scalability. In this paper, we present four implementations of K-means data clustering algorithm for different high performance computing platforms. These four implementations include a CUDA implementation for GPUs, a Mitrion C implementation for FPGAs, an MPI implementation for Beowulf compute clusters, and an OpenMP implementation for shared-memory machines. The comparative analyses of the cost of each platform, difficulty level of programming for each platform, and the performance of each implementation are presented.

Parallel K-Means Clustering for Brain Cancer Detection Using Hyperspectral Images

“…We also made comparisons with other recent works in the literature, that only in one case achieved results comparable but not better than ours, except for the FPGA solution proposed in [27]. However, FPGA memory constraint does not allow to process images with more than 17, 692 pixels.…”

“…A comparative analysis similar to the one we conducted is reported in [27]. Authors exploited GPUs, OpenMP, Message Passing Interface (MPI) and FPGAs.…”

“…Finally, concerning the FPGA implementation, experiments were reported only with a 17, 692 9-dimensionality dataset. Classification time is~100 ms, but, as stated in [27], the FPGA resources, especially memory banks, were not enough to process bigger datasets. The authors of this work do not use external DDR memory, therefore, the FPGA performance is limited due to this design choice.…”

“…Maximum Image Size Data Dimensionality Technology Speed-Up [18] 2,000,000 8 GPU NVIDIA GTX 280 220 [19] 500,000 2 GPU NVIDIA 9600 GT 14 [20] 1,000,000 2 GPU NVIDIA 8800 GTX 60 [21] 15,052,800 3 4 × GPU NVIDIA GTX 750Ti 60 [22] 1,000,000 32 GPU NVIDIA GTX 280 N. A. [23] 16,777,216 3 GPU NVIDIA Tesla C2050 25 [24] 245,057 4 GPU NVIDIA GeForce 210 386 [25] 500,000 16 GPU NVIDIA Quadro K5000 88 [26] N. A. N. A. GPU NVIDIA GTX 1080 18.5 [27] 20,000 10 2 × AMD Opteron quad-core 8 [27] 65,536 10 GPU NVIDIA Tesla 2050 60 [27] 17,692 9 Mitrion MVP FPGA Simulator N. A. Our work 264,408 128 GPU NVIDIA GTX 1060 126…”

Efficient fingerprint matching using GPU

“…Peralta et al [17] proposed a twolevel parallelised cluster framework that combined process-level and thread-level parallelisms. For large-scale automated FI system (AFIS), cluster-based solutions are expensive, as their cost depends on the number of high-end computers used in the cluster [18]. DSP-based accelerators offer limited parallelism and they may not be adequate for exploiting the maximum parallelism of minutiaebased fingerprint matching algorithms.…”