Jinkun Geng

Mr. Jinkun Geng

orcid

Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 November 2018 in CCF Transactions on Networking
Verified
Journal Article
Open access
PDF file
YCYang Cheng
Yang Cheng
Jinkun GengJinkun Geng
Jinkun Geng
YWYanshu Wang
Yanshu Wang
JLJunfeng Li
Junfeng Li
DLDan Li
Dan Li
JWJianping Wu
Jianping Wu
With the booming development of artificial intelligence (AI), a series of relevant applications are emerging and promoting an all-rounded reform of the industry. As the major technology of AI, machine learning (ML) shows great potential in solving network challenges. Network optimization, in return, brings significant performance gains for ML applications, in particular distributed machine learning. In this paper, we conduct a survey on combining ML technologies with network research.

19 June 2020 in IEEE/ACM Transactions on Networking
Verified
Journal Article
Closed access
SWSongtao Wang
Songtao Wang
DLDan Li
Dan Li
YCYang Cheng
Yang Cheng
Jinkun GengJinkun Geng
Jinkun Geng
YWYanshu Wang
Yanshu Wang
JWJianping Wu
Jianping Wu
In large-scale distributed machine learning (DML), the network performance between machines significantly impacts the speed of iterative training. In this paper we propose BML , a scalable, high-performance and fault-tolerant DML network architecture on top of Ethernet and commodity devices. BML builds on BCube topology, and runs a fully-distributed gradient synchronization algorithm. Compared to a Fat-Tree network with the same size, a BML network is expected to take much less time for gradient synchronization, for both low theoretical synchronization time and its benefit to RDMA transport. With server/link failures, the performance of BML degrades in a graceful way. Experiments of MNIST and VGG-19 benchmarks on a testbed with 9 dual-GPU servers show that, BML reduces the job completion time of DML training by up to 56.4% compared with Fat-Tree running state-of-the-art gradient synchronization algorithm.

8 October 2021 in IEEE/ACM Transactions on Networking
Verified
Journal Article
Closed access
SWShuai Wang
Shuai Wang
Jinkun GengJinkun Geng
Jinkun Geng
DLDan Li
Dan Li
To tackle the increasingly larger training data and models, researchers and engineers resort to multiple servers in a data center for distributed machine learning (DML). On one hand, DML enables us to leverage the computation power of multiple servers, which can effectively accelerate those computation-intensive tasks. On the other hand, DML also incurs significant communication cost due to parameter synchronization among these servers. In this paper, we want to explore the impact of synchronization topology, including both logical topology and physical topology, on the DML performance. First, we revisit the existing logical topologies, e.g., parameter server and ring allreduce, for parameter synchronization, and we find that these flat synchronization topologies is inefficient when running a large-scale DML training. Therefore, we propose a hierarchical parameter synchronization topology, called HiPS, which can achieve efficient parameter synchronization even on a large scale. Then, we compare two representative physical network topologies, namely, Fat-Tree and BCube. Based on our analyses, BCube has many advantages over Fat-Tree, e.g., higher bandwidth, better load balance, and lower hardware cost. The simulation results also show that BCube is more friendly to RDMA. Relying on the advantages of HiPS and BCube, the GST of “HiPS+BCube” is 12% ~ 70% lower than other combinations. Moreover, when the cluster size increases from 16 to 1024, the performance of “HiPS+BCube” only drops by 6.5%, while the performance of “Ring+BCube” drops by 44.6%. Hence, we believe “HiPS+BCube” is the optimal solution to benefit DML in large scale.

20 May 2024 in Journal of Systems Architecture
Verified
Journal Article
Closed access
RZRunhua Zhang
Runhua Zhang
HJHongxu Jiang
Hongxu Jiang
Jinkun GengJinkun Geng
Jinkun Geng
FTFangzheng Tian
Fangzheng Tian
YMYuhang Ma
Yuhang Ma
HWHaojie Wang
Haojie Wang

29 December 2020 in IEEE Transactions on Parallel and Distributed Systems
Verified
Journal Article
Closed access
YCYang Cheng
Yang Cheng
DLDan Li
Dan Li
ZGZhiyuan Guo
Zhiyuan Guo
BJBinyao Jiang
Binyao Jiang
Jinkun GengJinkun Geng
Jinkun Geng
JWJianping Wu
Jianping Wu
In this paper, we investigate the performance bottleneck of existing deep learning (DL) systems and propose DLBooster to improve the running efficiency of deploying DL applications on GPU clusters. At its core, DLBooster leverages two-level optimizations to speed up the end-to-end DL workflow. On the one hand, DLBooster selectively offloads some key decoding workloads to FPGAs to provide high-performance online data preprocessing services to the computing engine. On the other hand, DLBooster reorganizes the computational workloads of training neural networks with the back-propagation algorithm and schedules them according to their dependencies to improve the hardware utilization of GPUs at runtime. Based on our experiments, we prove that compared with baselines, DLBooster can improve the image processing throughput by 1.4× - 2.5× and reduce the processing latency by 1/3 in several real-world DL applications and datasets. Moreover, DLBooster consumes < 1 CPU core to manage FPGA devices at runtime, which is at least 90% less than the baselines in some cases.DLBooster shows its potential to accelerate DL workflows in the cloud.

4 August 2024
Verified
Conference Paper
Open access
MHMuhammad Haseeb
Muhammad Haseeb
Jinkun GengJinkun Geng
Jinkun Geng
UBUlysses Butler
Ulysses Butler
XHXiyu Hao
Xiyu Hao
DDDaniel Duclos-Cavalcanti
Daniel Duclos-Cavalcanti
ASAnirudh Sivaraman
Anirudh Sivaraman
There has been a growing interest from industry [9, 13, 14] and academia [8, 10] in migrating financial exchanges to the public cloud because of multiple benefits provided by the cloud, e.g., robust infrastructure, and potential cost savings [4]. However, migrating financial exchanges from on-prem clusters to the cloud poses several challenges.