• Proceedings of the Korea Information Processing Society Conference
• /
• 2011.11a
• /
• pp.718-721
• /
• 2011

웹 서비스를 대상으로 하는 DDoS(Distributed Denial-of-Service) 공격 또는 유해 트래픽 유입을 탐지 또는 차단하기 위한 목적으로 HTTP(Hypertext Transfer Protocol) 트래픽을 실시간으로 분석하는 기능은 거의 모든 네트워크 트래픽 보안 솔루션들이 탑재하고 있는 필수적인 요소이다. 하지만, HTTP 트래픽의 실시간 데이터 측정 양이 시간이 지날수록 기하급수적으로 증가함에 따라, HTTP 트래픽을 실시간 패킷 단위로 분석한다는 것에 대한 성능 부담감은 날로 커지고 있는 실정이다. 이제는 응용 어플리케이션 차원에서는 성능에 대한 부담감을 해소할 수 없기 때문에 고비용의 소프트웨어 가속기나 하드웨어에 의존적인 전용 장비를 탑재하여 해결하려는 시도가 대부분이다. 본 논문에서는 현재 대부분의 PC 에 탑재되어 있는 그래픽 카드의 GPU(Graphics Processing Units)를 범용적으로 활용하고자 하는 GPGPU(General-Purpose computation on Graphics Processing Units)의 연구에 힘입어, NVIDIA사의 CUDA(Compute Unified Device Architecture)를 사용하여 네트워크 트래픽에서 HTTP 패킷 추출성능을 응용 어플리케이션 차원에서 향상시켜 보고자 하였다. HTTP 패킷 추출 연산만을 기준으로 GPU 의 연산속도는 CPU 에 비해 10 배 이상의 높은 성능을 얻을 수 있었다.

• Journal of the Optical Society of Korea
• /
• v.15 no.4
• /
• pp.386-393
• /
• 2011

In this study, we report an effective k-domain linearization method with a pre-calibrated indexed look-up table. The method minimizes k-domain nonlinear characteristics of a swept source optical coherence tomography (SS-OCT) system by using two arrays, a sample position shift index and an intensity compensation array. Two arrays are generated from an interference pattern acquired by connecting a Fabry-Perot interferometer (FPI) and an optical spectrum analyzer (OSA) to the system. At real time imaging, the sample position is modified by location movement and intensity compensation with two arrays for linearity of wavenumber. As a result of evaluating point spread functions (PSFs), the signal to noise ratio (SNR) is increased by 9.7 dB. When applied to infrared (IR) sensing card imaging, the SNR is increased by 1.29 dB and the contrast noise ratio (CNR) value is increased by 1.44. The time required for the linearization and intensity compensation is 30 ms for a multi thread method using a central processing unit (CPU) compared to 0.8 ms for compute unified device architecture (CUDA) processing using a graphics processing unit (GPU). We verified that our linearization method is appropriate for applying real time imaging of SS-OCT.

• Journal of KIISE:Computer Systems and Theory
• /
• v.37 no.2
• /
• pp.66-75
• /
• 2010

This paper proposes an efficient method for design and implementation of the artificial intelligence (AI) of 'omok' game on the GPU. The proposed AI is designed on a cooperative structure using min-max game tree and genetic algorithm. Since the evaluation function needs intensive computation but is independently performed on a lot of candidates in the solution space, it is computed on the GPU in a massive parallel way. The implementation on NVIDIA CUDA and the experimental results show that it outperforms significantly over the CPU, in which parallel game tree and genetic algorithm on the GPU runs more than 400 times and 300 times faster than on the CPU. In the proposed cooperative AI, selective search using genetic algorithm is performed subsequently after the full search using game tree to search the solution space more efficiently as well as to avoid the thread overflow. Experimental results show that the proposed algorithm enhances the AI significantly and makes it run within the time limit given by the game's rule.