대규모 무선 센서네트워크를 위한 클러스터 기반 라우팅 기법

이윤석(Yunseok Rhee) 한국정보기술학회 2011 한국정보기술학회논문지 Vol.9 No.2

In this paper, we propose a cluster management scheme and a tree network of the cluster heads. Based on the scheme, we also present a routing scheme which supports the broadcast to the whole network as well as a data delivery to an arbitrary node from the sink node. Compared to the typical dissemination scheme, the proposed scheme enables the broadcast with reduction of about 43-50% in network traffic load, and thus significantly reduces the amount of energy consumption by 34-38%. Compared to AODV which supports a point-to-point delivery, in terms of both network traffic volume and delivery time, our scheme is also expected to show comparable performance as an established cluster structure is long sustained and the delivery data size increases.

작업영역의 동적 할당을 통한 고화질 애니메이션의 병렬 렌더링

이윤석(Yunseok Rhee) 한국컴퓨터정보학회 2008 韓國컴퓨터情報學會論文誌 Vol.13 No.1

고화질 입체 영상의 효과적인 재생을 위해 PC 클러스터를 활용한 여러 형태의 병렬화 기법이 제안되었지만, 영상을 구성하는 객체의 분포가 균일하지 않은 경우 충분한 성능을 발휘하지 못하였다. 본 연구에서는 POV-Ray 렌더러를 채택한 PC 클러스터 기반의 병렬 렌더링 시스템을 구축하고, 병렬화 성능을 높이기 위한 효과적인 부하 균형 기법을 개발하였다. 특히 애니메이션을 구성하는 연속 프레임 작업에서 프레임 간의 연관성(coherence)이 높다는 사실에 근거하여, 임의 프레임의 각 분할 영역에 소요된 계산량을 바탕으로 다음 프레임의 부하 분포를 예측하고 이에 맞게 각 프로세서의 작업 영역을 재조정하는 기법을 제안하였다. 제안 기법의 성능을 평가하기 위해, 충분하지는 않지만 2개의 실제 애니메이션 데이터에 대한 적용 결과, 정적 분할에 비해 약 40% 가량의 성능 향상을 보였다. 또한 다양한 부하 분포에 대한 각 기법의 성능을 추정하기 위해 수행한 모의실험에서, 정적 분할 기법에 대해 부하균형, 확장성 측면에서 우월한 것으로 예측되었다. Even though many studies on parallel rendering based on PC clusters have been done, most of those did not cope with non-uniform scenes, where locations of 3D models are biased. In this work, we have built a PC cluster system with POV-Ray, a free rendering software on the public domain, and developed an adaptive load balancing scheme to optimize the parallel efficiency. Especially, we noticed that a frame of 3D animation are closely coherent with adjacent frames, and thus we could estimate distribution of computation amount, based on the computation time of previous frame. The experimental results with 2 real animation data show that the proposed scheme reduces by 40% of execution time compared to the simple static partitioning scheme.

A Scalable Dissemination Network Architecture for Real-time Sensing Data Delivery

Yunseok Rhee(이윤석) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.2

When highly distributed sensor networks are fully realized, they will bring up new issues in delivering sensing data streams to data consumers in real-time. Such a data delivery scheme should be traffic efficient because the Internet can be easily inundated by a massive number of data streams. This paper proposes a scalable data dissemination network architecture (SDNA) that provides not only shared management and measurement of network information but also traffic efficient data delivery. The SDNA basically organizes a hierarchical cluster based on a virtual layer over the Internet. It can significantly reduce the amount of data traffic by establishing a commonly shared delivery path to multiple consumers on the cluster structure.

Energy Use Coordinator for Multiple Personal Sensor Devices

Yunseok Rhee(이윤석) 한국컴퓨터정보학회 2017 韓國컴퓨터情報學會論文誌 Vol.22 No.2

Useful continuous sensing applications are increasingly emerging as a new class of mobile applications. Meanwhile, open, multi-use sensor devices are newly adopted beyond smartphones, and provide huge opportunities to expand potential application categories. In this upcoming environment, uncoordinated use of sensor devices would cause severe imbalance in power consumption of devices, and thus result in early shutdown of some sensing applications depending on power-hungry devices. In this paper, we propose EnergyCordy, a novel inter-device energy use coordination system; with a system-wide holistic view, it coordinates the energy use of concurrent sensing applications over multiple sensor devices. As its key approach, we propose a relaxed sensor association; it decouples the energy use of an application from specific sensor devices leveraging multiple context inference alternatives, allowing flexible energy coordination at runtime. We demonstrated the effectiveness of EnergyCordy by developing multiple example applications over custom-designed wearable senor devices. We show that EnergyCordy effectively coordinates the power usage of concurrent sensing applications over multiple devices and prevent undesired early shutdown of applications.

QoS-aware Data Delivery Infrastructure for IoT Computing Environments

Yunseok Rhee(이윤석) 한국디지털콘텐츠학회 2018 한국디지털콘텐츠학회논문지 Vol.19 No.2

파일 접근 행태의 LSTM 학습을 활용한 악성 코드 탐지 기법

이윤석(Yunseok Rhee) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.2

In this paper, we propose a malicious code detection method based on deep learning of files access behaviors. Unlike the static code analysis, this method runs actual malicious codes on an isolated virtual machine and collects file system call patterns and behavior. In particular, for this study, we employ the FUSE-based user space file system, and it enables collecting all the file access data without modifying the kernel. In order to effectively train long input patterns, the LSTM network, an improved RNN model, is adopted. As a data set, five kinds of malicious code samples used in the previous study are selected and about 4,500 malicious and positive code patterns are generated. The classification performance is 92.2%, which is 3.8 ~ 44.7% better than the existing machine learning.

GPU 기반 병렬 Ant Colony System의 구현과 성능 분석

이윤석(Yunseok Rhee) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.2

Ant Colony System is a metaheuristic technique that seeks a near optimal solution for a combinatorial problem with high computational complexity. Effective parallelization of algorithms is required due to the enormous amount of computation of ACS. In this paper, we propose an ACS algorithm that makes the most of data parallelism provided on GPU, and analyze its performance with the Traveling Salesman Problem to In particular, in order to increase data parallelism, the number of parallel threads and thread blocks were maximized, and their cocurrent continuous memory accesses were fully utilized. In addition, for effective TSP resolution, a process of searching for optimal solutions using proximity by distance between nodes is proposed. This experiment was implemented using Nvidia Titan RTX GPU and Inteli9-9900K and showed the effectiveness of performance improvement through experiments on public TSPLIB data.

장소 방문 기록을 활용한 행동 예측 모델

이윤석(Yunseok Rhee) 한국정보기술학회 2013 한국정보기술학회논문지 Vol.11 No.10

More interest has been focused on activity-aware applications recently. In this paper, we propose an activity prediction model based on influential context history which includes individual’s visit place, visit time, duration, and transportation mode. The influential contexts can be attained from people’s mobility information. Especially, it is well known that the mobility information inherently shows a strong activity causality and activity sequence pattern. Thus, based on the causality, we design a Bayesian network (BN) model to predict next activities and to detect activity sequence patterns. To improve the initial simple BN system, we also build a hierarchical Bayesian network, which performs better in terms of resource use and accuracy. This model supports more precise prediction on the place compared previous researches mostly based on simple temporal GPS data. From the experiment for simple places using ATUS data, the proposed system shows higher accuracy than previous works. It also shows quite promising result for commercial complex places where the place information is not definitely identified.

그리드 환경에서 어플리케이션의 특성을 고려한 자원선택 시스템의 설계

정평재(Pyoung-Jae Jung),이윤석(Yunseok Rhee) 한국정보과학회 2004 한국정보과학회 학술발표논문집 Vol.31 No.1A

본 연구에서는 웹서비스 기반 그리드 환경에서 효과적인 자원선택을 할 수 있는 자원 선택 프레임웍을 설계 구현하였다. 이 프레임웍을 활용하여, 어플리케이션이 요구하는 자원 명세와 자원 공급자의 자원공급 정책을 기술할 수 있도록 지원하고, 이를 토대로 다양한 자원 선택알고리즘을 개발할 수 있을 것으로 기대된다. 또한 본 연구에서 구현한 프로토타입 시스템을 소규모의 그리드 환경에 적용해 얻은 실험 결과를 통해, 그리드 상에서 어플리케이션의 성질을 고려한 자원 선택이 충분한 효과를 나타냄을 확인할 수 있었다.

디렉토리를 이용한 캐쉬 일관성 유지 기법에서 무효화 힌트를 이용한 읽기 접근 시간 감소

오승택(Seungtaek Oh),이윤석(Yunseok Rhee),맹승렬(Seung Ryoul Maeng),이준원(Joonwon Lee) 한국정보과학회 2000 정보과학회논문지 : 시스템 및 이론 Vol.27 No.4

대규모 분산 공유메모리 다중처리기는 공유메모리 접근 지연시간이 크다는 약점을 지니고 있다. 이러한 다중처리기에서 모든 메모리 요청이 홈노드를 통해 이루어지는 디렉토리 기반의 캐쉬 일관성 유지 기법의 사용은 메모리 접근 지연시간을 더욱 크게 하는 요인으로 작용한다. 뿐만 아니라 메모리 접근 지연시간은 시스템의 규모가 커질수록 전체 성능에 중요한 요소로 작용하므로, 대규모 시스템에서 이를 줄이기 위해서 많은 연구들이 있었다. 본 논문에서는 메모리 읽기 지연시간을 줄이는 새로운 캐쉬 일관성 유지 기법을 제안한다. 제안된 기법은 무효화힌트를 이용하여 구현되었다. 무효화힌트는 어떤 노드가 전에 캐쉬블록을 무효화 시켰는가에 관한 정보이며, 메모리블록이 필요한 노드는 이 정보를 이용하여 홈노드의 도움 없이 직접 메모리 요청을 할 수 있다. 제안된 프로토콜의 성능을 측정하기 위하여 모의실험을 하였다. 모의실험 결과는 제안된 프로토콜에서 읽기 지연시간이 감소하는 것을 나타낸다. Large scale shared memory multiprocessors have suffered from large access latency to shared memory. The large latency partially stems from a feature of directory-based cache coherence schemes which require a shared memory access to be serviced at a home node of the memory block. The home visit results in three or more hops traversal for a memory read access. The traversal becomes much longer as a system scales up. In this paper, we propose a new cache coherence scheme that reduces read access latency. The proposed scheme exploits ideas of invalid hint. Invalid hint for a cache block means which node has invalidated the cache block before. Thus a read access request can be directly sent to and serviced by the node (called owner) without help of a home node. Execution-driven simulation is employed to evaluate performance of the proposed scheme. The simulation results show that read access latency and execution time are reduced.