On-Line Prediction of Nonstationary Variable-Bit-Rate Video Traffic

Sungjoo Kang,Seongjin Lee,Youjip Won,Byeongchan Seong IEEE 2010 IEEE transactions on signal processing Vol.58 No.3

<P>In this paper, we propose a model-based bandwidth prediction scheme for variable-bit-rate (VBR) video traffic with regular group of pictures (GOP) pattern. Multiplicative ARIMA (autoregressive integrated moving-average) process called GOP ARIMA (ARIMA for GOP) is used as a base stochastic model, which consists of two key ingredients: prediction and model validity check. For traffic prediction, we deploy a Kalman filter over GOP ARIMA model, and confidence interval analysis for validity determination. The GOP ARIMA mPodel explicitly models inter and intra-GOP frame size correlations and the Kalman filter-based prediction maintains ?state? across the prediction rounds. Synergy of the two successfully addresses a number of challenging issues, such as a unified framework for frame type dependent prediction, accurate prediction, and robustness against noise. With few exceptions, a single video session consists of several scenes whose bandwidth process may exhibit different stochastic nature, which hinders recursive adjustment of parameters in Kalman filter, because its stochastic model structure is fixed at its deployment. To effectively address this issue, the proposed prediction scheme harbors a statistical hypothesis test in the prediction framework. By formulating the confidence interval of a prediction in terms of Kalman filter components, it not only <I>predicts</I> the frame size but also <I>determines</I> validity of the stochastic model. Based upon the results of the model validity check, the proposed prediction scheme updates the structures of the underlying GOP ARIMA model. We perform a comprehensive performance study using publicly available MPEG-2 and MPEG-4 traces. We compare the prediction accuracy of four different prediction schemes. In all traces, the proposed model yields superior prediction accuracy than the other prediction schemes. We show that confidence interval analysis effectively detects the structural changes in the sample sequence and that properly updating the model results in more accurate prediction. However, model update requires a certain length of observation period, e.g., 60 frames (2 s). Due to this learning overhead, the advantage of model update becomes less significant when scene length is short. Through queueing simulation, we examine the effect of prediction accuracy over user perceivable QoS. The proposed bandwidth prediction scheme allocates less 50% of the queue(buffer) compared to the other bandwidth prediction schemes, but still yields better packet loss behavior.</P>

Non-stationary VBR 트래픽을 위한 GOP ARIMA 기반 대역폭 예측 기법

강성주(Sungjoo Kang),원유집(Youjip Won) 한국통신학회 2004 한국통신학회 학술대회논문집 Vol.2004 No.11

Non-stationary VBR 트래픽을 위한 동적 데이타 크기 예측 알고리즘

강성주(Sungjoo Kang),원유집(Youjip Won),성병찬(Byeongchan Seong) 한국정보과학회 2007 정보과학회논문지 : 정보통신 Vol.34 No.3

본 논문에서는 VBR(Variable-Bit-Rate) 트래픽의 비선형적이고 버스티한 특성을 모델화 한 GOP ARIMA(ARIMA for Group Of Pictures) 모델을 칼만 필터 알고리즘을 이용하여 실시간으로 예측하는 기법을 제안한다. 칼만 필터를 이용한 예측 기법은 GOP ARIMA의 상태공간 모델링 과정과 향후 N초 간의 트래픽을 예측하는 과정으로 구성된다. 실험을 위해 GOP의 크기가 각각 15인 세 가지 종류의 MPEG VBR 트래픽(뉴스, 드라마, 스포츠)을 제작하였고, 칼만 필터를 이용한 세 가지 종류의 트래픽의 예측 결과를 선형 예측법과 이중 지수 평활법을 이용해 예측한 결과와 비교해 예측 성능이 상대적으로 우수함을 확인할 수 있었다. 또한 예측값에 신뢰 구간을 설정하는 신뢰 구간 분석법을 통해 트래픽 관점에서 장면 변화를 예측하는 방법을 제시하였다. 본 논문의 칼만 필터 기반의 예측 알고리즘은 MPEG 기반 VBR 트래픽을 비롯한 기타 인터넷 트래픽을 실시간으로 예측하는 방법과 이를 이용해 인터넷 서버의 설계 및 자원 할당 정책 등을 위한 트래픽 엔지니어링 연구에 기여할 수 있을 것이다. In this paper, we develop the model based prediction algorithm for Variable-Bit-Rate(VBR) video traffic with regular Group of Picture(GOP) pattern. We use multiplicative ARIMA process called GOP ARIMA (ARIMA for Group Of Pictures) as a base stochastic model. Kalman Filter based prediction algorithm consists of two process: GOP ARIMA modeling and prediction. In performance study, we produce three video traces (news, drama, sports) and we compare the accuracy of three different prediction schemes: Kalman Filter based prediction, linear prediction, and double exponential smoothing. The proposed prediction algorithm yields superior prediction accuracy than the other two. We also show that confidence interval analysis can effectively detect scene changes of the sample video sequence. The Kalman filter based prediction algorithm proposed in this work makes significant contributions to various aspects of network traffic engineering and resource allocation.

스마트-커넥티드 제품의 생명주기 관리를 위한 설계-제조-서비스 연동 아키텍처 연구

강성주(Sungjoo Kang),전재호(Jaeho Jeon),김현수(Hyeon-Soo Kim),전인걸(Ingeol Chun) 한국정보과학회 2015 한국정보과학회 학술발표논문집 Vol.2015 No.6

고신뢰 사이버-물리 무기체계 획득을 위한 LVC 연동 개발 프레임워크

강성주(Sungjoo Kang),김민조(Minjo Kim),박정민(Jungmin Park),전인걸(Ingeol Chun),김원태(Wontae Kim) 한국통신학회 2013 韓國通信學會論文誌 Vol.38 No.12(융합기술)

본 논문에서는 사이버-물리 시스템의 모델링 및 시뮬레이션 도구인 EcoSuite를 기반으로 지능화, 복잡화되고 있는 사이버-물리 시스템 형태의 무기체계를 개발 및 시험하는 프레임워크를 제시한다. EcoPOD를 이용한 무기체계의 모델링과 타 무기체계 모델들과 연동하여 구성 (Constructive) 시뮬레이션 환경 기반의 전장을 제공하는 EcoSIM이 소개된다. LVC 연동 개발을 위해서는 시뮬레이션 모델 연동 구조와 연동 기술에 순응(compliant)하는 인터페이스 기술, 그리고 시뮬레이션 모델과 실제 시스템(Live), 그리고 사용자와 상호작용하는 시스템(Virtual)의 연계 기술의 개발이 필요하다. 본 논문에서는 LVC 연동 개발을 통한 모델의 검증 및 시스템의 기능을 시험하는 아키텍처와 적용 사례가 제시된다. In this paper, we present a development framework for acquiring intelligent but complex cyber-physical weapon systems based on modeling and simulation development tools for cyber-physical systems, EcoSUITE. We introduce EcoPOD that models weapon systems and EcoSIM that provides constructive simulation environment for interoperating the weapon model to be developed with other weapon models. To develop cyber-physical weapon system based on LVC interoperation, an interoperation architecture and an interface technique for a live and a virtual system that is compliant with the interoperation architecture. By expanding EcoSuite, we provide LVC-based development framework for interoperating a real system, a human-interactive interface system, and simulation models and validate it with a case study.

객체 지향 DBMS에서 질의 재수행 과정이 없는 뷰 메카니즘의 설계

강성주(Sungjoo Kang),박지숙(Jisook Park),이석호(Sukho Lee) 한국정보과학회 1995 한국정보과학회 학술발표논문집 Vol.22 No.2A

OODB에서 뷰를 제공할 때 RDB에서의 저장 뷰 방법을 그대로 적용할 경우 객체 중첩의 문제가 발생하게 된다. 본 논문에서는 뷰에 저장된 저장 객체의 OID와 애트리뷰트 경로로 저장 객체의 데이타에 접근할 수 있는 방법을 제시하였으며, 이를 통해 뷰 실행시 매번 질의를 재수행하지 않으면서 관계 데이타베이스 시스템에서 쓰이는 저장 뷰 방법이 객체 지향 모델에서 가지는 문제점을 극복하였다. 특히 이 방법은 뷰 정의문의 조건절에서 지정한 애트리뷰트에 대한 변경이 거의 없는 환경에서 효율적이다.

MPEG VBR트래픽을 위한 GOP ARIMA기반 대역폭 예측기법

강성주(Sungjoo Kang),원유집(Youjip Won) 대한전기학회 2004 대한전기학회 학술대회 논문집 Vol.2004 No.11

A Study on Fake Iris Detection based on the Reflectance of the Iris to the Sclera for Iris Recognition

Sungjoo Lee,Kang Ryoung Park,Jaihie Kim 대한전자공학회 2005 ITC-CSCC :International Technical Conference on Ci Vol.2005 No.4

Controllable Nondegenerate p-Type Doping of Tungsten Diselenide by Octadecyltrichlorosilane

Kang, Dong-Ho,Shim, Jaewoo,Jang, Sung Kyu,Jeon, Jeaho,Jeon, Min Hwan,Yeom, Geun Young,Jung, Woo-Shik,Jang, Yun Hee,Lee, Sungjoo,Park, Jin-Hong American Chemical Society 2015 ACS NANO Vol.9 No.2

<P>Despite heightened interest in 2D transition-metal dichalcogenide (TMD) doping methods for future layered semiconductor devices, most doping research is currently limited to molybdenum disulfide (MoS<SUB>2</SUB>), which is generally used for n-channel 2D transistors. In addition, previously reported TMD doping techniques result in only high-level doping concentrations (degenerate) in which TMD materials behave as near-metallic layers. Here, we demonstrate a controllable nondegenerate p-type doping (p-doping) technique on tungsten diselenide (WSe<SUB>2</SUB>) for p-channel 2D transistors by adjusting the concentration of octadecyltrichlorosilane (OTS). This p-doping phenomenon originates from the methyl (−CH<SUB>3</SUB>) functional groups in OTS, which exhibit a positive pole and consequently reduce the electron carrier density in WSe<SUB>2</SUB>. The controlled p-doping levels are between 2.1 × 10<SUP>11</SUP> and 5.2 × 10<SUP>11</SUP> cm<SUP>–2</SUP> in the nondegenerate regime, where the performance parameters of WSe<SUB>2</SUB>-based electronic and optoelectronic devices can be properly designed or optimized (threshold voltage↑, on-/off-currents↑, field-effect mobility↑, photoresponsivity↓, and detectivity↓ as the doping level increases). The p-doping effect provided by OTS is sustained in ambient air for a long time showing small changes in the device performance (18–34% loss of Δ<I>V</I><SUB>TH</SUB> initially achieved by OTS doping for 60 h). Furthermore, performance degradation is almost completely recovered by additional thermal annealing at 120 °C. Through Raman spectroscopy and electrical/optical measurements, we have also confirmed that the OTS doping phenomenon is independent of the thickness of the WSe<SUB>2</SUB> films. We expect that our controllable p-doping method will make it possible to successfully integrate future layered semiconductor devices.</P><P><B>Graphic Abstract</B> <IMG SRC='http://pubs.acs.org/appl/literatum/publisher/achs/journals/content/ancac3/2015/ancac3.2015.9.issue-2/nn5074435/production/images/medium/nn-2014-074435_0007.gif'></P><P><A href='http://pubs.acs.org/doi/suppl/10.1021/nn5074435'>ACS Electronic Supporting Info</A></P>

Reinforcement Learning-Assisted Garbage Collection to Mitigate Long-Tail Latency in SSD

Kang, Wonkyung,Shin, Dongkun,Yoo, Sungjoo Association for Computing Machinery 2017 ACM transactions on embedded computing systems Vol.16 No.s5

<P>NAND flash memory is widely used in various systems, ranging from real-time embedded systems to enterprise server systems. Because the flash memory has erase-before-write characteristics, we need flash-memory management methods, i.e., address translation and garbage collection. In particular, garbage collection (GC) incurs long-tail latency, e.g., 100 times higher latency than the average latency at the 99th percentile. Thus, real-time and quality-critical systems fail to meet the given requirements such as deadline and QoS constraints. In this study, we propose a novel method of GC based on reinforcement learning. The objective is to reduce the long-tail latency by exploiting the idle time in the storage system. To improve the efficiency of the reinforcement learning-assisted GC scheme, we present new optimization methods that exploit fine-grained GC to further reduce the long-tail latency. The experimental results with real workloads show that our technique significantly reduces the long-tail latency by 29-36% at the 99.99th percentile compared to state-of-the-art schemes.</P>