윤대건
Daegun Yoon
kljp at ajou.ac.kr
Research interests
Machine Learning, Parallel Algorithm, Distributed System
Introduction
Daegun Yoon is a Ph.D. candidate at the Department of Artificial Intelligence, Ajou University. He received the B.S. in the Department of Software, Ajou University, in 2018.
Publications
2024
23.
Yoon, Daegun; Oh, Sang-yoon
Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning🌏 InternationalConference
The 24th IEEE/ACM international Symposium on Cluster, Cloud and Internet Computing (CCGrid 2024), 2024.
@conference{icpp2024yoon,
title = {Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning},
author = {Daegun Yoon and Sang-yoon Oh},
url = {https://arxiv.org/abs/2402.13781},
year = {2024},
date = {2024-02-13},
booktitle = {The 24th IEEE/ACM international Symposium on Cluster, Cloud and Internet Computing (CCGrid 2024)},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
2023
22.
Yoon, Daegun; Oh, Sangyoon
MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training🌏 InternationalConference
30th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC 2023), 2023.
@conference{nokey,
title = {MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training},
author = {Daegun Yoon and Sangyoon Oh},
url = {https://arxiv.org/abs/2310.00967},
year = {2023},
date = {2023-10-02},
urldate = {2023-10-02},
booktitle = {30th IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC 2023)},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
21.
Yoon, Daegun; Oh, Sangyoon
DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification🌏 InternationalConference
International Conference on Parallel Processing (ICPP) 2023, 2023.
@conference{nokey,
title = {DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification},
author = {Daegun Yoon and Sangyoon Oh},
url = {https://arxiv.org/abs/2307.03500},
year = {2023},
date = {2023-08-07},
urldate = {2023-08-07},
booktitle = {International Conference on Parallel Processing (ICPP) 2023},
abstract = {Gradient sparsification is a widely adopted solution for reducing
the excessive communication traffic in distributed deep learning.
However, most existing gradient sparsifiers have relatively poor
scalability because of considerable computational cost of gradient
selection and/or increased communication traffic owing to gradient
build-up. To address these challenges, we propose a novel gradient
sparsification scheme, DEFT, that partitions the gradient selection
task into sub tasks and distributes them to workers. DEFT differs
from existing sparsifiers, wherein every worker selects gradients
among all gradients. Consequently, the computational cost can
be reduced as the number of workers increases. Moreover, gradient build-up can be eliminated because DEFT allows workers to
select gradients in partitions that are non-intersecting (between
workers). Therefore, even if the number of workers increases, the
communication traffic can be maintained as per user requirement.
To avoid the loss of significance of gradient selection, DEFT
selects more gradients in the layers that have a larger gradient
norm than the other layers. Because every layer has a different
computational load, DEFT allocates layers to workers using a binpacking algorithm to maintain a balanced load of gradient selection
between workers. In our empirical evaluation, DEFT shows a significant improvement in training performance in terms of speed
in gradient selection over existing sparsifiers while achieving high
convergence performance.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Gradient sparsification is a widely adopted solution for reducing
the excessive communication traffic in distributed deep learning.
However, most existing gradient sparsifiers have relatively poor
scalability because of considerable computational cost of gradient
selection and/or increased communication traffic owing to gradient
build-up. To address these challenges, we propose a novel gradient
sparsification scheme, DEFT, that partitions the gradient selection
task into sub tasks and distributes them to workers. DEFT differs
from existing sparsifiers, wherein every worker selects gradients
among all gradients. Consequently, the computational cost can
be reduced as the number of workers increases. Moreover, gradient build-up can be eliminated because DEFT allows workers to
select gradients in partitions that are non-intersecting (between
workers). Therefore, even if the number of workers increases, the
communication traffic can be maintained as per user requirement.
To avoid the loss of significance of gradient selection, DEFT
selects more gradients in the layers that have a larger gradient
norm than the other layers. Because every layer has a different
computational load, DEFT allocates layers to workers using a binpacking algorithm to maintain a balanced load of gradient selection
between workers. In our empirical evaluation, DEFT shows a significant improvement in training performance in terms of speed
in gradient selection over existing sparsifiers while achieving high
convergence performance.
the excessive communication traffic in distributed deep learning.
However, most existing gradient sparsifiers have relatively poor
scalability because of considerable computational cost of gradient
selection and/or increased communication traffic owing to gradient
build-up. To address these challenges, we propose a novel gradient
sparsification scheme, DEFT, that partitions the gradient selection
task into sub tasks and distributes them to workers. DEFT differs
from existing sparsifiers, wherein every worker selects gradients
among all gradients. Consequently, the computational cost can
be reduced as the number of workers increases. Moreover, gradient build-up can be eliminated because DEFT allows workers to
select gradients in partitions that are non-intersecting (between
workers). Therefore, even if the number of workers increases, the
communication traffic can be maintained as per user requirement.
To avoid the loss of significance of gradient selection, DEFT
selects more gradients in the layers that have a larger gradient
norm than the other layers. Because every layer has a different
computational load, DEFT allocates layers to workers using a binpacking algorithm to maintain a balanced load of gradient selection
between workers. In our empirical evaluation, DEFT shows a significant improvement in training performance in terms of speed
in gradient selection over existing sparsifiers while achieving high
convergence performance.
20.
유미리,; 윤대건,; 오상윤,
연합학습 기법들의 성능평가를 지원하는 이기종 기반의 실험 플랫폼 설계🇰🇷 DomesticConference
2023년도 한국통신학회 하계종합학술발표회 , 한국통신학회 2023.
@conference{연합학습기법들의성능평가를지원하는이기종기반의실험플랫폼설계,
title = {연합학습 기법들의 성능평가를 지원하는 이기종 기반의 실험 플랫폼 설계},
author = {유미리 and 윤대건 and 오상윤},
url = {https://www.dbpia.co.kr/journal/articleDetail?nodeId=NODE11487802},
year = {2023},
date = {2023-06-21},
urldate = {2023-06-21},
booktitle = {2023년도 한국통신학회 하계종합학술발표회 },
organization = { 한국통신학회},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
19.
Lee, Seungjun; Yu, Miri; Yoon, Daegun; Oh, Sangyoon
Can hierarchical client clustering mitigate the data heterogeneity effect in federated learning?🌏 InternationalConference
2023 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 2023, ISBN: 979-8-3503-1200-3.
@conference{nokey,
title = {Can hierarchical client clustering mitigate the data heterogeneity effect in federated learning?},
author = {Seungjun Lee and Miri Yu and Daegun Yoon and Sangyoon Oh},
url = {10.1109/IPDPSW59300.2023.00134},
doi = {10.1109/IPDPSW59300.2023.00134},
isbn = {979-8-3503-1200-3},
year = {2023},
date = {2023-05-15},
urldate = {2023-05-15},
booktitle = {2023 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)},
abstract = {Federated learning (FL) was proposed for training a deep neural network model using millions of user data. The technique has attracted considerable attention owing to its privacy-preserving characteristic. However, two major challenges exist. The first is the limitation of simultaneously participating clients. If the number of clients increases, the single parameter server easily becomes a bottleneck and is prone to have stragglers. The second is data heterogeneity, which adversely affects the accuracy of the global model. Because data should remain at user devices to preserve privacy, we cannot use data shuffling, which is used to homogenize training data in traditional distributed deep learning. We propose a client clustering and model aggregation method, CCFed, to increase the number of simultaneously participating clients and mitigate the data heterogeneity problem. CCFed improves the learning performance using set partition modeling to let data be evenly distributed between clusters and mitigate the effect of a non-IID environment. Experiments show that we can achieve a 2.7-14% higher accuracy using CCFed compared with FedAvg, where CCFed requires approximately 50% less number of rounds compared with FedAvg training on benchmark datasets.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Federated learning (FL) was proposed for training a deep neural network model using millions of user data. The technique has attracted considerable attention owing to its privacy-preserving characteristic. However, two major challenges exist. The first is the limitation of simultaneously participating clients. If the number of clients increases, the single parameter server easily becomes a bottleneck and is prone to have stragglers. The second is data heterogeneity, which adversely affects the accuracy of the global model. Because data should remain at user devices to preserve privacy, we cannot use data shuffling, which is used to homogenize training data in traditional distributed deep learning. We propose a client clustering and model aggregation method, CCFed, to increase the number of simultaneously participating clients and mitigate the data heterogeneity problem. CCFed improves the learning performance using set partition modeling to let data be evenly distributed between clusters and mitigate the effect of a non-IID environment. Experiments show that we can achieve a 2.7-14% higher accuracy using CCFed compared with FedAvg, where CCFed requires approximately 50% less number of rounds compared with FedAvg training on benchmark datasets.
18.
최지헌,; 유미리,; 윤대건,; 오상윤,
연합학습에서의 보안 취약점 분석🇰🇷 DomesticConference
2023년도 한국통신학회 동계종합학술발표회 논문집 , vol. 80, 한국통신학회 2023, ISSN: 2383-8302.
@conference{최지헌2023연합학습에서의,
title = {연합학습에서의 보안 취약점 분석},
author = {최지헌 and 유미리 and 윤대건 and 오상윤},
url = {https://www.dbpia.co.kr/pdf/pdfView.do?nodeId=NODE11227811},
issn = {2383-8302},
year = {2023},
date = {2023-02-28},
urldate = {2023-02-28},
booktitle = {2023년도 한국통신학회 동계종합학술발표회 논문집
},
volume = {80},
pages = {1201-1202},
organization = {한국통신학회},
abstract = {개인 데이터에 대한 프라이버시 침해 없이 분산 기계학습을 구현하기 위해 연합학습이 제안되었다. 기존 연합학습 기법의 개선을 통해 정확도향상 및 수렴속도 향상을 목표로 하는 새로운 기법들이 등장하고 있어서, 이에 대한 보안 가이드라인이 필요한 상황이다. 본 논문에서는연합학습 구조의 특징으로 나타나는 보안 취약점을 공격형태 별로 구분하고 이에 대한 대응방안을 고찰한다.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
개인 데이터에 대한 프라이버시 침해 없이 분산 기계학습을 구현하기 위해 연합학습이 제안되었다. 기존 연합학습 기법의 개선을 통해 정확도향상 및 수렴속도 향상을 목표로 하는 새로운 기법들이 등장하고 있어서, 이에 대한 보안 가이드라인이 필요한 상황이다. 본 논문에서는연합학습 구조의 특징으로 나타나는 보안 취약점을 공격형태 별로 구분하고 이에 대한 대응방안을 고찰한다.
17.
Yoon, Daegun; Jeong, Minjoong; Oh, Sangyoon
SAGE: toward on-the-fly gradient compression ratio scaling🌏 InternationalJournal Article
In: The Journal of Supercomputing, pp. 1–23, 2023.
@article{yoon2023sage,
title = {SAGE: toward on-the-fly gradient compression ratio scaling},
author = {Daegun Yoon and Minjoong Jeong and Sangyoon Oh},
url = {https://link.springer.com/article/10.1007/s11227-023-05120-7},
doi = {https://doi.org/10.1007/s11227-023-05120-7},
year = {2023},
date = {2023-02-25},
urldate = {2023-02-25},
journal = {The Journal of Supercomputing},
pages = {1--23},
abstract = {Gradient sparsification is widely adopted in distributed training; however, it suffers from a trade-off between computation and communication. The prevalent Top-k sparsifier has a hard constraint on computational overhead while achieving the desired gradient compression ratio. Conversely, the hard-threshold sparsifier eliminates computational constraints but fail to achieve the targeted compression ratio. Motivated by this tradeoff, we designed a novel threshold-based sparsifier called SAGE, which achieves a compression ratio close to that of the Top-k sparsifier with negligible computational overhead. SAGE scales the compression ratio by deriving an adjustable threshold based on each iteration’s heuristics. Experimental results show that SAGE achieves a compression ratio closer to the desired ratio than a hard-threshold sparsifier without exacerbating the accuracy of model training. In terms of computation time for gradient selection, SAGE achieves a speedup of up to 23.62×
over the Top-k sparsifier.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Gradient sparsification is widely adopted in distributed training; however, it suffers from a trade-off between computation and communication. The prevalent Top-k sparsifier has a hard constraint on computational overhead while achieving the desired gradient compression ratio. Conversely, the hard-threshold sparsifier eliminates computational constraints but fail to achieve the targeted compression ratio. Motivated by this tradeoff, we designed a novel threshold-based sparsifier called SAGE, which achieves a compression ratio close to that of the Top-k sparsifier with negligible computational overhead. SAGE scales the compression ratio by deriving an adjustable threshold based on each iteration’s heuristics. Experimental results show that SAGE achieves a compression ratio closer to the desired ratio than a hard-threshold sparsifier without exacerbating the accuracy of model training. In terms of computation time for gradient selection, SAGE achieves a speedup of up to 23.62×
over the Top-k sparsifier.
over the Top-k sparsifier.
2022
16.
Yoon, Daegun; Jeong, Minjoong; Oh, Sangyoon
WAVE: designing a heuristics-based three-way breadth-first search on GPUs🌏 InternationalJournal Article
In: The Journal of Supercomputing, 2022, (2).
@article{Yoon2022WAVE,
title = {WAVE: designing a heuristics-based three-way breadth-first search on GPUs},
author = {Daegun Yoon and Minjoong Jeong and Sangyoon Oh},
doi = {10.1007/s11227-022-04934-1},
year = {2022},
date = {2022-11-17},
urldate = {2022-11-17},
journal = {The Journal of Supercomputing},
abstract = {Breadth-first search (BFS) is a building block for improving the performance of many iterative graph algorithms. In addition to conventional BFS (push), a novel method that traverses a graph in the reverse direction (pull) has emerged and gained popularity because of its enhanced processing performance. Several frameworks have recently used a hybrid approach known as direction-optimizing BFS, which utilizes both directions. However, these frameworks are mostly interested in optimizing the procedure in each direction, instead of designing sophisticated methods for determining the appropriate direction between push and pull at each iteration. Owing to the lack of in-depth discussion on this decision, state-of-the-art direction-optimizing BFS algorithms cannot demonstrate their comprehensive performance despite improvements in the design of each direction because they select ineffective directions at each iteration. We identified that the current frameworks suffer from high computational overheads for their decisions and make decisions that are overfitted to several graph datasets used for tuning their direction selection process. Based on observations from state-of-the-art limitations, we designed a direction-optimizing method for BFS called WAVE. WAVE minimizes the computational overhead to near zero and makes more appropriate direction selection decisions than the state-of-the-art heuristics based on the characteristics extracted from the input graph dataset. In our experiments on 20 graph benchmarks, WAVE achieved speedups of up to 4.95×, 5.79×, 46.49×, and 149.67× over Enterprise, Gunrock, Tigr, and CuSha, respectively.},
note = {2},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Breadth-first search (BFS) is a building block for improving the performance of many iterative graph algorithms. In addition to conventional BFS (push), a novel method that traverses a graph in the reverse direction (pull) has emerged and gained popularity because of its enhanced processing performance. Several frameworks have recently used a hybrid approach known as direction-optimizing BFS, which utilizes both directions. However, these frameworks are mostly interested in optimizing the procedure in each direction, instead of designing sophisticated methods for determining the appropriate direction between push and pull at each iteration. Owing to the lack of in-depth discussion on this decision, state-of-the-art direction-optimizing BFS algorithms cannot demonstrate their comprehensive performance despite improvements in the design of each direction because they select ineffective directions at each iteration. We identified that the current frameworks suffer from high computational overheads for their decisions and make decisions that are overfitted to several graph datasets used for tuning their direction selection process. Based on observations from state-of-the-art limitations, we designed a direction-optimizing method for BFS called WAVE. WAVE minimizes the computational overhead to near zero and makes more appropriate direction selection decisions than the state-of-the-art heuristics based on the characteristics extracted from the input graph dataset. In our experiments on 20 graph benchmarks, WAVE achieved speedups of up to 4.95×, 5.79×, 46.49×, and 149.67× over Enterprise, Gunrock, Tigr, and CuSha, respectively.
15.
윤대건,; 노병희,; 오상윤,
전술망의 라우팅 성능 개선을 위한 성능 지표 분석 기반 정책 엔진 설계🇰🇷 DomesticJournal Article
In: 한국통신학회 논문지, vol. 47, iss. 9, no. 9, pp. 1353-1359, 2022.
@article{윤대건2022전술망,
title = {전술망의 라우팅 성능 개선을 위한 성능 지표 분석 기반 정책 엔진 설계},
author = {윤대건 and 노병희 and 오상윤},
url = {https://www.kci.go.kr/kciportal/ci/sereArticleSearch/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877193},
year = {2022},
date = {2022-10-31},
urldate = {2022-10-31},
journal = {한국통신학회 논문지},
volume = {47},
number = {9},
issue = {9},
pages = {1353-1359},
abstract = {컴퓨팅 관련 기술 발달에 따라 군 작전 수행에서 발생하는 데이터의 규모가 매우 커지고 있으며, 이에 따라 이를 처리하기 위한 군 전술망의 성능 향상에 대한 요구 또한 점점 늘어나고 있다. 군 전술망의 특성 상 다양한 장비로 구성된 네트워크를 활용해야 하며, 이러한 상황에서 민간에서 활발히 적용되는 Software-Defined Network (SDN) 기술을 적용한다면 장비를 제공 벤더로부터 자유로운 손쉬운 네트워크 관리가 가능하다. 본 논문에서는SDN 기반 네트워크 환경에서 패킷 전송 성능 향상을 목적으로 하는 네트워크 정책 엔진 구조 설계를 소개한다.
정책 엔진은 Flow table의 Flow들이 나타내는 라우팅 경로를 수정하도록 하는 알고리즘을 포함하며 성능 개선 여부는 본 연구에서 정의한 종합 성능 지표를 통해 판단한다. 추후 본 연구에서 제안하는 전술망 라우팅 성능 개량을 위한 성능 지표 분석 기반 정책 엔진 기반의 소프트웨어를 실제 네트워크 운용 상황에 적용하고, 네트워크 성능 향상을 검증하도록 할 계획이다.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
컴퓨팅 관련 기술 발달에 따라 군 작전 수행에서 발생하는 데이터의 규모가 매우 커지고 있으며, 이에 따라 이를 처리하기 위한 군 전술망의 성능 향상에 대한 요구 또한 점점 늘어나고 있다. 군 전술망의 특성 상 다양한 장비로 구성된 네트워크를 활용해야 하며, 이러한 상황에서 민간에서 활발히 적용되는 Software-Defined Network (SDN) 기술을 적용한다면 장비를 제공 벤더로부터 자유로운 손쉬운 네트워크 관리가 가능하다. 본 논문에서는SDN 기반 네트워크 환경에서 패킷 전송 성능 향상을 목적으로 하는 네트워크 정책 엔진 구조 설계를 소개한다.
정책 엔진은 Flow table의 Flow들이 나타내는 라우팅 경로를 수정하도록 하는 알고리즘을 포함하며 성능 개선 여부는 본 연구에서 정의한 종합 성능 지표를 통해 판단한다. 추후 본 연구에서 제안하는 전술망 라우팅 성능 개량을 위한 성능 지표 분석 기반 정책 엔진 기반의 소프트웨어를 실제 네트워크 운용 상황에 적용하고, 네트워크 성능 향상을 검증하도록 할 계획이다.
정책 엔진은 Flow table의 Flow들이 나타내는 라우팅 경로를 수정하도록 하는 알고리즘을 포함하며 성능 개선 여부는 본 연구에서 정의한 종합 성능 지표를 통해 판단한다. 추후 본 연구에서 제안하는 전술망 라우팅 성능 개량을 위한 성능 지표 분석 기반 정책 엔진 기반의 소프트웨어를 실제 네트워크 운용 상황에 적용하고, 네트워크 성능 향상을 검증하도록 할 계획이다.
14.
Yoon, Daegun; Oh, Sangyoon
The 8th International Conference on Next Generation Computing (ICNGC) 2022, 2022.
@conference{yoon2022empirical,
title = {Empirical Analysis on Top-k Gradient Sparsification for Distributed Deep Learning in a Supercomputing Environment},
author = {Daegun Yoon and Sangyoon Oh},
doi = {10.48550/arXiv.2209.08497},
year = {2022},
date = {2022-09-19},
booktitle = {The 8th International Conference on Next Generation Computing (ICNGC) 2022},
abstract = {To train deep learning models faster, distributed training on multiple GPUs is the very popular scheme in recent years. However, the communication bandwidth is still a major bottleneck of training performance. To improve overall training performance, recent works have proposed gradient sparsification methods that reduce the communication traffic significantly. Most of them require gradient sorting to select meaningful gradients such as Top-k gradient sparsification (Top-k SGD). However, Top-k SGD has a limit to increase the speed up overall training performance because gradient sorting is significantly inefficient on GPUs. In this paper, we conduct experiments that show the inefficiency of Top-k SGD and provide the insight of the low performance. Based on observations from our empirical analysis, we plan to yield a high performance gradient sparsification method as a future work. },
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
To train deep learning models faster, distributed training on multiple GPUs is the very popular scheme in recent years. However, the communication bandwidth is still a major bottleneck of training performance. To improve overall training performance, recent works have proposed gradient sparsification methods that reduce the communication traffic significantly. Most of them require gradient sorting to select meaningful gradients such as Top-k gradient sparsification (Top-k SGD). However, Top-k SGD has a limit to increase the speed up overall training performance because gradient sorting is significantly inefficient on GPUs. In this paper, we conduct experiments that show the inefficiency of Top-k SGD and provide the insight of the low performance. Based on observations from our empirical analysis, we plan to yield a high performance gradient sparsification method as a future work.
13.
이승준,; 윤대건,; 오상윤,
SDN 정책엔진의 사용자 모듈을 위한 분석 요청 정의 언어🇰🇷 DomesticJournal Article
In: 한국통신학회 논문지, vol. 47, no. 9, pp. 1360-1369, 2022.
@article{이승준2022SDN,
title = {SDN 정책엔진의 사용자 모듈을 위한 분석 요청 정의 언어},
author = {이승준 and 윤대건 and 오상윤},
url = {https://www.kci.go.kr/kciportal/ci/sereArticleSearch/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART002877194},
year = {2022},
date = {2022-09-01},
urldate = {2022-09-01},
journal = {한국통신학회 논문지},
volume = {47},
number = {9},
pages = {1360-1369},
abstract = {현대전에서 작전 수행은 네트워크 중심전의 양상을 띄고 있으며, 이에 따라 군 전술망 자원을 효과적으로 사용하기 위한 여러 연구가 진행되고 있다. 단일 연구 결과가 아닌 여러 연구의 결과를 복합적으로 적용했을 때의 효과를 분석하기 위한 노력의 일환으로 통합 테스트베드가 구축되고, 여기에서 여러 네트워크 알고리즘을 동시에 수행하고 성능을 분석하기 위한 정책 엔진도 설계되었다. 하지만 이종 네트워크 환경에서는 사용자들이 요구하는 서른 다른 데이터 구조의 성능 지표와 이를 처리할 각 알고리즘의 서로 다른 실행 환경에 적응적으로 대응하기 어려운 문제가 있었다. 이에 본 연구에서는 프로그래밍 언어와 실행 환경 등 특정 기술에 종속되지 않는 정책 엔진을 위한 XML 기반의 인터페이스 포맷을 정의하고 그 스키마를 제안한다. 제안된 스키마를 사용하여 메시지는 특정 프로그래밍 언어에 종속되지 않고 인코딩과 디코딩을 할 수 있으며 Open Container Initiative 표준을 기반으로실행 환경을 정의하는 컨테이너를 기술할 수 있다.
In modern warfare environment, the well defined networks becomes important to the operations. Thus, researchers study on how to use the military tactical network resources effectively. To analyze effectiveness of the results from multiple studies together, an integrated testbed is critical as well as the design and the implementation of a policy engine that performs multiple network algorithms and analyze performance simultaneously. However, when the network environment is heterogeneous, it is hard to respond adaptively to the performance indicators of the different data structures and the different execution environments of each user algorithm. To address this issue, we propose an XML-based interface format and its schema for the policy engine, which is independent from specific technologies such as programming languages and execution environments. A message from and to the policy engine and the testbed can be encoded and decoded regardless of the programming language. Furthermore, it can describe containers of the execution environment based on the Open Container Initiative standard.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
현대전에서 작전 수행은 네트워크 중심전의 양상을 띄고 있으며, 이에 따라 군 전술망 자원을 효과적으로 사용하기 위한 여러 연구가 진행되고 있다. 단일 연구 결과가 아닌 여러 연구의 결과를 복합적으로 적용했을 때의 효과를 분석하기 위한 노력의 일환으로 통합 테스트베드가 구축되고, 여기에서 여러 네트워크 알고리즘을 동시에 수행하고 성능을 분석하기 위한 정책 엔진도 설계되었다. 하지만 이종 네트워크 환경에서는 사용자들이 요구하는 서른 다른 데이터 구조의 성능 지표와 이를 처리할 각 알고리즘의 서로 다른 실행 환경에 적응적으로 대응하기 어려운 문제가 있었다. 이에 본 연구에서는 프로그래밍 언어와 실행 환경 등 특정 기술에 종속되지 않는 정책 엔진을 위한 XML 기반의 인터페이스 포맷을 정의하고 그 스키마를 제안한다. 제안된 스키마를 사용하여 메시지는 특정 프로그래밍 언어에 종속되지 않고 인코딩과 디코딩을 할 수 있으며 Open Container Initiative 표준을 기반으로실행 환경을 정의하는 컨테이너를 기술할 수 있다.
In modern warfare environment, the well defined networks becomes important to the operations. Thus, researchers study on how to use the military tactical network resources effectively. To analyze effectiveness of the results from multiple studies together, an integrated testbed is critical as well as the design and the implementation of a policy engine that performs multiple network algorithms and analyze performance simultaneously. However, when the network environment is heterogeneous, it is hard to respond adaptively to the performance indicators of the different data structures and the different execution environments of each user algorithm. To address this issue, we propose an XML-based interface format and its schema for the policy engine, which is independent from specific technologies such as programming languages and execution environments. A message from and to the policy engine and the testbed can be encoded and decoded regardless of the programming language. Furthermore, it can describe containers of the execution environment based on the Open Container Initiative standard.
In modern warfare environment, the well defined networks becomes important to the operations. Thus, researchers study on how to use the military tactical network resources effectively. To analyze effectiveness of the results from multiple studies together, an integrated testbed is critical as well as the design and the implementation of a policy engine that performs multiple network algorithms and analyze performance simultaneously. However, when the network environment is heterogeneous, it is hard to respond adaptively to the performance indicators of the different data structures and the different execution environments of each user algorithm. To address this issue, we propose an XML-based interface format and its schema for the policy engine, which is independent from specific technologies such as programming languages and execution environments. A message from and to the policy engine and the testbed can be encoded and decoded regardless of the programming language. Furthermore, it can describe containers of the execution environment based on the Open Container Initiative standard.
12.
Park, Juwon; Yoon, Daegun; Yeo, Sangho; Oh, Sangyoon
AMBLE: Adjusting Mini-Batch and Local Epoch for Federated Learning with Heterogeneous Devices🌏 InternationalJournal Article
In: Journal of Parallel and Distributed Computing, 2022, ISSN: 0743-7315.
@article{Juwon2022AMBLE,
title = {AMBLE: Adjusting Mini-Batch and Local Epoch for Federated Learning with Heterogeneous Devices},
author = {Juwon Park and Daegun Yoon and Sangho Yeo and Sangyoon Oh},
url = {https://www.sciencedirect.com/science/article/pii/S0743731522001757},
doi = {https://doi.org/10.1016/j.jpdc.2022.07.009},
issn = {0743-7315},
year = {2022},
date = {2022-07-21},
urldate = {2022-07-21},
journal = {Journal of Parallel and Distributed Computing},
abstract = {As data privacy becomes increasingly important, federated learning applied to the training of deep learning models while ensuring the data privacy of devices is entering the spotlight. Federated learning makes it possible to process all data at once while processing data independently from various devices without collecting distributed local data in a central server. However, there are still challenges to overcome for the system of devices in federated learning such as communication overheads and the heterogeneity of the system. In this paper, we propose the Adjusting Mini-Batch and Local Epoch (AMBLE) approach, which adaptively adjusts the local mini-batch and local epoch size for heterogeneous devices in federated learning and updates the parameters synchronously. With AMBLE, we enhance the computational efficiency by removing stragglers and scaling the local learning rate to improve the model convergence rate and accuracy. We verify that federated learning with AMBLE is a stably trained model with a faster convergence speed and higher accuracy than FedAvg and adaptive batch size scheme for both identically and independently distributed (IID) and non-IID cases.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
As data privacy becomes increasingly important, federated learning applied to the training of deep learning models while ensuring the data privacy of devices is entering the spotlight. Federated learning makes it possible to process all data at once while processing data independently from various devices without collecting distributed local data in a central server. However, there are still challenges to overcome for the system of devices in federated learning such as communication overheads and the heterogeneity of the system. In this paper, we propose the Adjusting Mini-Batch and Local Epoch (AMBLE) approach, which adaptively adjusts the local mini-batch and local epoch size for heterogeneous devices in federated learning and updates the parameters synchronously. With AMBLE, we enhance the computational efficiency by removing stragglers and scaling the local learning rate to improve the model convergence rate and accuracy. We verify that federated learning with AMBLE is a stably trained model with a faster convergence speed and higher accuracy than FedAvg and adaptive batch size scheme for both identically and independently distributed (IID) and non-IID cases.
11.
Yoon, Daegun; Oh, Sangyoon
SURF: Direction-Optimizing Breadth-First Search Using Workload State on GPUs🌏 InternationalJournal Article
In: Sensors, vol. 22, no. 13, pp. 4899, 2022.
@article{yoon2022surf,
title = {SURF: Direction-Optimizing Breadth-First Search Using Workload State on GPUs},
author = {Daegun Yoon and Sangyoon Oh},
url = {https://www.mdpi.com/1424-8220/22/13/4899},
doi = {https://doi.org/10.3390/s22134899},
year = {2022},
date = {2022-06-29},
urldate = {2022-01-01},
journal = {Sensors},
volume = {22},
number = {13},
pages = {4899},
publisher = {Multidisciplinary Digital Publishing Institute},
abstract = { Graph data structures have been used in a wide range of applications including scientific and social network applications. Engineers and scientists analyze graph data to discover knowledge and insights by using various graph algorithms. A breadth-first search (BFS) is one of the fundamental building blocks of complex graph algorithms and its implementation is included in graph libraries for large-scale graph processing. In this paper, we propose a novel direction selection method, SURF (Selecting directions Upon Recent workload of Frontiers) to enhance the performance of BFS on GPU. A direction optimization that selects the proper traversal direction of a BFS execution between the push and pull phases is crucial to the performance as well as for efficient handling of the varying workloads of the frontiers. However, existing works select the direction using condition statements based on predefined thresholds without considering the changing workload state. To solve this drawback, we define several metrics that describe the state of the workload and analyze their impact on the BFS performance. To show that SURF selects the appropriate direction, we implement the direction selection method with a deep neural network model that adopts those metrics as the input features. Experimental results indicate that SURF achieves a higher direction prediction accuracy and reduced execution time in comparison with existing state-of-the-art methods that support a direction-optimizing BFS. SURF yields up to a 5.62×},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Graph data structures have been used in a wide range of applications including scientific and social network applications. Engineers and scientists analyze graph data to discover knowledge and insights by using various graph algorithms. A breadth-first search (BFS) is one of the fundamental building blocks of complex graph algorithms and its implementation is included in graph libraries for large-scale graph processing. In this paper, we propose a novel direction selection method, SURF (Selecting directions Upon Recent workload of Frontiers) to enhance the performance of BFS on GPU. A direction optimization that selects the proper traversal direction of a BFS execution between the push and pull phases is crucial to the performance as well as for efficient handling of the varying workloads of the frontiers. However, existing works select the direction using condition statements based on predefined thresholds without considering the changing workload state. To solve this drawback, we define several metrics that describe the state of the workload and analyze their impact on the BFS performance. To show that SURF selects the appropriate direction, we implement the direction selection method with a deep neural network model that adopts those metrics as the input features. Experimental results indicate that SURF achieves a higher direction prediction accuracy and reduced execution time in comparison with existing state-of-the-art methods that support a direction-optimizing BFS. SURF yields up to a 5.62×
10.
정현석,; 유미리,; 윤대건,; 이승준,; 오상윤,
재난 대응 기계학습 모델의 Data Drift 문제에 대한 MLOps 기반 대응 기법🇰🇷 DomesticConference
2022 한국차세대컴퓨팅학회 춘계학술대회, 한국차세대컴퓨팅학회, 2022.
@conference{정현석2022재난,
title = {재난 대응 기계학습 모델의 Data Drift 문제에 대한 MLOps 기반 대응 기법},
author = {정현석 and 유미리 and 윤대건 and 이승준 and 오상윤},
url = {https://www.earticle.net/Article/A412404},
year = {2022},
date = {2022-05-21},
urldate = {2022-05-21},
booktitle = {2022 한국차세대컴퓨팅학회 춘계학술대회},
pages = {pp.473-476},
publisher = {한국차세대컴퓨팅학회},
abstract = {기계학습에서 Data Drift는 정확도에 큰 영향을 주는 중요한 문제이며, 재난 대응과 같이 모델의 잘못된 예측 피해가 큰 분야에서 더 중요하다. 본 논문에서는 재난 분야 Data Drift 문제에 대해 MLOps를 이용하여 모델의 재학습을 효과적으로 수행할 수 있는 방안으로 MLOps 기법과 툴들을 사용하는 것을 제안하고, Kaggle 데이터와 MLFlow를 기반으로 정확도 실험을 수행하여 주장을 검증하였다.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
기계학습에서 Data Drift는 정확도에 큰 영향을 주는 중요한 문제이며, 재난 대응과 같이 모델의 잘못된 예측 피해가 큰 분야에서 더 중요하다. 본 논문에서는 재난 분야 Data Drift 문제에 대해 MLOps를 이용하여 모델의 재학습을 효과적으로 수행할 수 있는 방안으로 MLOps 기법과 툴들을 사용하는 것을 제안하고, Kaggle 데이터와 MLFlow를 기반으로 정확도 실험을 수행하여 주장을 검증하였다.
2021
9.
Lee, Seungjun; Yoon, Daegun; Yeo, Sangho; Oh, Sangyoon
Mitigating Cold Start Problem in Serverless Computing with Function Fusion🌏 InternationalJournal Article
In: Sensors, vol. 21, no. 24, 2021, ISSN: 1424-8220.
@article{s21248416,
title = {Mitigating Cold Start Problem in Serverless Computing with Function Fusion},
author = {Seungjun Lee and Daegun Yoon and Sangho Yeo and Sangyoon Oh},
url = {https://www.mdpi.com/1424-8220/21/24/8416},
doi = {10.3390/s21248416},
issn = {1424-8220},
year = {2021},
date = {2021-12-23},
urldate = {2021-12-16},
journal = {Sensors},
volume = {21},
number = {24},
abstract = {As Artificial Intelligence (AI) is becoming ubiquitous in many applications, serverless computing is also emerging as a building block for developing cloud-based AI services. Serverless computing has received much interest because of its simplicity, scalability, and resource efficiency. However, due to the trade-off with resource efficiency, serverless computing suffers from the cold start problem, that is, a latency between a request arrival and function execution. The cold start problem significantly influences the overall response time of workflow that consists of functions because the cold start may occur in every function within the workflow. Function fusion can be one of the solutions to mitigate the cold start latency of a workflow. If two functions are fused into a single function, the cold start of the second function is removed; however, if parallel functions are fused, the workflow response time can be increased because the parallel functions run sequentially even if the cold start latency is reduced. This study presents an approach to mitigate the cold start latency of a workflow using function fusion while considering a parallel run. First, we identify three latencies that affect response time, present a workflow response time model considering the latency, and efficiently find a fusion solution that can optimize the response time on the cold start. Our method shows a response time of 28%–86% of the response time of the original workflow in five workflows.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
As Artificial Intelligence (AI) is becoming ubiquitous in many applications, serverless computing is also emerging as a building block for developing cloud-based AI services. Serverless computing has received much interest because of its simplicity, scalability, and resource efficiency. However, due to the trade-off with resource efficiency, serverless computing suffers from the cold start problem, that is, a latency between a request arrival and function execution. The cold start problem significantly influences the overall response time of workflow that consists of functions because the cold start may occur in every function within the workflow. Function fusion can be one of the solutions to mitigate the cold start latency of a workflow. If two functions are fused into a single function, the cold start of the second function is removed; however, if parallel functions are fused, the workflow response time can be increased because the parallel functions run sequentially even if the cold start latency is reduced. This study presents an approach to mitigate the cold start latency of a workflow using function fusion while considering a parallel run. First, we identify three latencies that affect response time, present a workflow response time model considering the latency, and efficiently find a fusion solution that can optimize the response time on the cold start. Our method shows a response time of 28%–86% of the response time of the original workflow in five workflows.
8.
Yoon, Daegun; Oh, Sangyoon
Traversing Large Road Networks on GPUs with Breadth-First Search🌏 InternationalConference 📃 In press
The 7th International Conference on Next Generation Computing, 2021.
@conference{Yoon2021Traversing,
title = {Traversing Large Road Networks on GPUs with Breadth-First Search},
author = {Daegun Yoon and Sangyoon Oh},
year = {2021},
date = {2021-11-05},
urldate = {2021-11-05},
booktitle = {The 7th International Conference on Next Generation Computing},
journal = {The 7th International Conference on Next Generation Computing 2021},
abstract = {Breadth-first search (BFS) is one of the most used graph kernels, and substantially affects the overall performance when processing various graphs. Since graph data are frequently used in real life for example road networks in navigation systems, high performance graph processing becomes more critical. In this study, we aim to process BFS algorithm efficiently on road network data. We propose BARON, a BFS algorithm that copes with road networks. To accelerate graph traversal, BARON reduce the occurrence of branch and memory divergences by exploiting warp-cooperative work sharing and atomic operations. With this design approach, BARON outperforms the other BFS kernels of state-of-the-art graph processing frameworks executed stably on the latest GPU architectures. For various graphs, BARON yields speedups of up to 2.88x and 5.43x over Gunrock and CuSha, respectively.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Breadth-first search (BFS) is one of the most used graph kernels, and substantially affects the overall performance when processing various graphs. Since graph data are frequently used in real life for example road networks in navigation systems, high performance graph processing becomes more critical. In this study, we aim to process BFS algorithm efficiently on road network data. We propose BARON, a BFS algorithm that copes with road networks. To accelerate graph traversal, BARON reduce the occurrence of branch and memory divergences by exploiting warp-cooperative work sharing and atomic operations. With this design approach, BARON outperforms the other BFS kernels of state-of-the-art graph processing frameworks executed stably on the latest GPU architectures. For various graphs, BARON yields speedups of up to 2.88x and 5.43x over Gunrock and CuSha, respectively.
7.
Lee, Seungjun; Yoon, Daegun; Oh, Sangyoon
Imitation learning for VM placement problem using demonstration data generated by heuristics🌏 InternationalConference 📃 In press
17th Int. Conference on Data Science (ICDATA’21), 2021.
@conference{lee2021imitation,
title = {Imitation learning for VM placement problem using demonstration data generated by heuristics},
author = {Seungjun Lee and Daegun Yoon and Sangyoon Oh},
url = {https://youtu.be/CmG3E1rWroQ},
year = {2021},
date = {2021-07-26},
urldate = {2021-07-26},
booktitle = {17th Int. Conference on Data Science (ICDATA’21)},
abstract = {Data centers are key components of cloud computing to run virtual machines. For saving the cost to operate data centers, it is important to decide how to allocate each virtual machine to a certain physical machine. Because the virtual machine placement problem is NP-Hard, there are many heuristics to obtain near-optimal solutions as quickly as possible. The reinforcement learning technique can be applied for virtual machine placement problem. However, if the problem size gets bigger, the convergence speed of reinforcement learning gets slower. The possible solution is that the agent imitates the behavior of given demonstration, called imitation learning. In this paper, we propose a method combining reinforcement learning with imitation learning. In our proposed approach, demonstration data is generated by simple heuristics not human experts.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Data centers are key components of cloud computing to run virtual machines. For saving the cost to operate data centers, it is important to decide how to allocate each virtual machine to a certain physical machine. Because the virtual machine placement problem is NP-Hard, there are many heuristics to obtain near-optimal solutions as quickly as possible. The reinforcement learning technique can be applied for virtual machine placement problem. However, if the problem size gets bigger, the convergence speed of reinforcement learning gets slower. The possible solution is that the agent imitates the behavior of given demonstration, called imitation learning. In this paper, we propose a method combining reinforcement learning with imitation learning. In our proposed approach, demonstration data is generated by simple heuristics not human experts.
6.
윤대건,; 노병희,; 오상윤,
전술망 성능 개량을 위한 정책 엔진 인터페이스 설계🇰🇷 DomesticConference 📃 In press
2021 한국군사과학기술학회 종합학술대회, 2021.
@conference{윤대건2021전술망,
title = {전술망 성능 개량을 위한 정책 엔진 인터페이스 설계},
author = {윤대건 and 노병희 and 오상윤},
year = {2021},
date = {2021-06-10},
booktitle = {2021 한국군사과학기술학회 종합학술대회},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
5.
Yoon, Daegun; Li, Zhetao; Oh, Sangyoon
Balanced content space partitioning for pub/sub: a study on impact of varying partitioning granularity🌏 InternationalJournal Article
In: The Journal of Supercomputing, pp. 1–27, 2021.
@article{yoon2021balanced,
title = {Balanced content space partitioning for pub/sub: a study on impact of varying partitioning granularity},
author = {Daegun Yoon and Zhetao Li and Sangyoon Oh},
url = {https://link.springer.com/article/10.1007/s11227-021-03821-5},
year = {2021},
date = {2021-01-01},
journal = {The Journal of Supercomputing},
pages = {1--27},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2020
4.
Yoon, Daegun; Park, Gyudong; Oh, Sangyoon
Exploring a system architecture of content-based publish/subscribe system for efficient on-the-fly data dissemination🌏 InternationalJournal Article
In: Concurrency and Computation: Practice and Experience, pp. e6090, 2020.
@article{yoon2020exploring,
title = {Exploring a system architecture of content-based publish/subscribe system for efficient on-the-fly data dissemination},
author = {Daegun Yoon and Gyudong Park and Sangyoon Oh},
url = {https://onlinelibrary.wiley.com/doi/full/10.1002/cpe.6090},
year = {2020},
date = {2020-01-01},
journal = {Concurrency and Computation: Practice and Experience},
pages = {e6090},
publisher = {Wiley Online Library},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
3.
Yoon, Daegun; Park, Gyudong; Oh, Sangyoon
CPartition: a Correlation-Based Space Partitioning for Content-Based Publish/Subscribe Systems with Skewed Workload🌏 InternationalConference
2020 IEEE International Conference on Big Data and Smart Computing (BigComp), IEEE 2020, ISBN: 978-1-7281-6034-4.
@conference{yoon2020cpartition,
title = {CPartition: a Correlation-Based Space Partitioning for Content-Based Publish/Subscribe Systems with Skewed Workload},
author = {Daegun Yoon and Gyudong Park and Sangyoon Oh},
url = {https://ieeexplore.ieee.org/abstract/document/9070640},
doi = {10.1109/BigComp48618.2020.00-46},
isbn = {978-1-7281-6034-4},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {2020 IEEE International Conference on Big Data and Smart Computing (BigComp)},
pages = {377--384},
organization = {IEEE},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
2.
윤대건,; 오상윤,
Software-Defined Network 에서의 Conflict Resolution 을 위한 정책엔진 구조 및 전략 분석🇰🇷 DomesticConference
한국통신학회 학술대회논문집, 2020.
@conference{윤대건2020software,
title = {Software-Defined Network 에서의 Conflict Resolution 을 위한 정책엔진 구조 및 전략 분석},
author = {윤대건 and 오상윤},
url = {https://www.dbpia.co.kr/pdf/pdfView.do?nodeId=NODE10498659&mark=0&useDate=&ipRange=N&accessgl=Y&language=ko_KR&hasTopBanner=false},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {한국통신학회 학술대회논문집},
journal = {한국통신학회 학술대회논문집},
pages = {566--567},
abstract = {본 논문은 Software-Defined Network 에서 추가되는 정책들이 기존의 정책들과 일으키는 충돌을 해결하기 위해 존재하는 정책엔진의 새로운 설계를 위해 분석한 기존 연구 내용들을 소개하고 분석한 내용들을 기반으로 정책엔진의 conflict resolution 전략을 제안한다. 본 논문에서는 SDN 에 새로운 정책이 추가되는 경우 발생할 수 있는 conflict 를 감지한 후 해결하기 위해, 정책엔진이 conflict detector 와 conflict handler 로 구성되는 구조를 가정한다. Conflict detector 는 새로 추가되는 정책이 기존의 정책들과 충돌을 일으키는지 감지하고 conflict handler 는 conflict resolution 을 통해 문제가 되는 정책을 삭제하는 역할을 한다. 본 논문에서는 conflict handler 가 Recency 중심 전략과 Priority 중심 전략을 사용하여 문제가 되는 정책을 삭제하는 방안에 대해서 분석한 결과를 소개한다},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
본 논문은 Software-Defined Network 에서 추가되는 정책들이 기존의 정책들과 일으키는 충돌을 해결하기 위해 존재하는 정책엔진의 새로운 설계를 위해 분석한 기존 연구 내용들을 소개하고 분석한 내용들을 기반으로 정책엔진의 conflict resolution 전략을 제안한다. 본 논문에서는 SDN 에 새로운 정책이 추가되는 경우 발생할 수 있는 conflict 를 감지한 후 해결하기 위해, 정책엔진이 conflict detector 와 conflict handler 로 구성되는 구조를 가정한다. Conflict detector 는 새로 추가되는 정책이 기존의 정책들과 충돌을 일으키는지 감지하고 conflict handler 는 conflict resolution 을 통해 문제가 되는 정책을 삭제하는 역할을 한다. 본 논문에서는 conflict handler 가 Recency 중심 전략과 Priority 중심 전략을 사용하여 문제가 되는 정책을 삭제하는 방안에 대해서 분석한 결과를 소개한다
2018
1.
윤대건,; 오상윤,
동종 운영체제 환경에서의 가상 머신 마이그레이션 성능 분석🇰🇷 DomesticConference
한국정보과학회 KCC 학술대회, 2018.
@conference{윤대건2018동종,
title = {동종 운영체제 환경에서의 가상 머신 마이그레이션 성능 분석},
author = {윤대건 and 오상윤},
year = {2018},
date = {2018-01-01},
urldate = {2018-01-01},
booktitle = {한국정보과학회 KCC 학술대회},
pages = {47--49},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}