Proceedings of the Workshop on Network Meets AI & ML最新文献

Challenges in Using ML for Networking Research: How to Label If You Must 使用机器学习进行网络研究的挑战:如果必须的话如何标记

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-14 DOI: 10.1145/3405671.3405812

Yukhe Lavinia, Ramakrishnan Durairajan, R. Rejaie, W. Willinger

引用次数: 3

Pitfalls of data-driven networking: A case study of latent causal confounders in video streaming 数据驱动网络的陷阱:视频流中潜在因果混淆的案例研究

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-14 DOI: 10.1145/3405671.3405815

P. C. Sruthi, Sanjay G. Rao, Bruno Ribeiro

引用次数: 10

Neural Packet Routing 神经分组路由

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-14 DOI: 10.1145/3405671.3405813

Shihan Xiao, Haiyan Mao, Bo-Xi Wu, Wenjie Liu, Fenglin Li

引用次数: 10

A Deep Learning Approach for IP Hijack Detection Based on ASN Embedding 基于ASN嵌入的IP劫持检测深度学习方法

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-14 DOI: 10.1145/3405671.3405814

T. Shapira, Y. Shavitt

引用次数: 10

An Adaptive Tree Algorithm to Approach Collision-Free Transmission in Slotted ALOHA 开槽ALOHA无碰撞传输的自适应树算法

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-14 DOI: 10.1145/3405671.3405817

Molly Zhang, L. D. Alfaro, J. Garcia-Luna-Aceves

引用次数: 1

DeepBGP

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-10 DOI: 10.1145/3405671.3405816

M. Bahnasy, Fenglin Li, Shihan Xiao, Xiangle Cheng

{"title":"DeepBGP","authors":"M. Bahnasy, Fenglin Li, Shihan Xiao, Xiangle Cheng","doi":"10.1145/3405671.3405816","DOIUrl":"https://doi.org/10.1145/3405671.3405816","url":null,"abstract":"Border Gateway Protocol (BGP) is the standard inter-domain routing protocol that is used to exchange reachability information among Wide Area Networks (WANs). BGP is a policy-based routing protocol that introduces a lot of flexibility. However, this flexibility increases the configuration complexity. In this research, we introduce DeepBGP as a neural network-based system that synthesizes network configuration given a high-level operator intent. We adopt Graph Neural Network (GNN) to represent network topology and generate partial network configuration. A validation unit is then used to calculate a reward based on which an Evolution Strategies (ES) optimizer updates neural network parameters. Since ES does not require backpropagation, they provide a significant reduction in calculation time. Further, the recent advances in deep learning with strong hardware acceleration and the parallelization capabilities offered by ES provide great potential in scaling the proposed solution to larger topologies. We demonstrate experimentally that DeepBGP can generate a network-wide configuration for both Huawei and Cisco devices while fulfilling operator requirements. We also show how Deep-BGP scales when the network size increases, and how hardware acceleration could improve the scalability of the system.","PeriodicalId":254313,"journal":{"name":"Proceedings of the Workshop on Network Meets AI & ML","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124770883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 9

Proceedings of the Workshop on Network Meets AI & ML 网络与人工智能和机器学习研讨会论文集

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-10 DOI: 10.1145/3405671

引用次数: 1

SmartEntry

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-10 DOI: 10.1145/3405671.3405809

Junjie Zhang, Zehua Guo, Minghao Ye, H. J. Chao

{"title":"SmartEntry","authors":"Junjie Zhang, Zehua Guo, Minghao Ye, H. J. Chao","doi":"10.1145/3405671.3405809","DOIUrl":"https://doi.org/10.1145/3405671.3405809","url":null,"abstract":"Traffic Engineering (TE) has been used by Internet service providers to improve their network performance and provide better service quality to users. While flow-based TE is an alternative, destination-based TE is a more readily deployed solution. This is because destination-based forwarding is ubiquitously supported by today's routers. A challenge faced by state-of-the-art destination-based TE solutions is considerable time taken by a centralized controller to update traffic split ratios for each entry of the forwarding table of each router. This could impose a fundamental limitation on how responsively the network can react to dynamic changes of traffic demands. In this paper, we propose SmartEntry, a destination-based routing solution coupled with Reinforcement Learning (RL) to reduce the number of the forwarding entries that need to be updated to respond to dynamic change of traffic demands. SmartEntry forwards majority traffic on Equal-Cost Multi-Path (ECMP) and redistributes a small portion of traffic using our proposed RL algorithm. SmartEntry adopts Linear Programming (LP) to produce reward signals. This RL + LP combined approach turns out to be surprisingly effective. We evaluate SmartEntry by conducting extensive experiments on different network topologies with both real and synthesized traffic. The simulation results show that SmartEntry achieves near-optimal performance with a saving of 90% forwarding entry updates, and generalizes well to unseen traffic matrices.","PeriodicalId":254313,"journal":{"name":"Proceedings of the Workshop on Network Meets AI & ML","volume":" 48","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120832132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 9

SAM: Self-Attention based Deep Learning Method for Online Traffic Classification 基于自注意的深度学习在线流量分类方法

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-08-10 DOI: 10.1145/3405671.3405811

Guorui Xie, Qing Li, Yong Jiang, Tao Dai, Gengbiao Shen, Rui Li, R. Sinnott, Shutao Xia

引用次数: 14

Is Network the Bottleneck of Distributed Training? 网络是分布式训练的瓶颈吗?

Proceedings of the Workshop on Network Meets AI & ML Pub Date : 2020-06-17 DOI: 10.1145/3405671.3405810

Zhen Zhang, Chaokun Chang, Haibin Lin, Yida Wang, R. Arora, Xin Jin

{"title":"Is Network the Bottleneck of Distributed Training?","authors":"Zhen Zhang, Chaokun Chang, Haibin Lin, Yida Wang, R. Arora, Xin Jin","doi":"10.1145/3405671.3405810","DOIUrl":"https://doi.org/10.1145/3405671.3405810","url":null,"abstract":"Recently there has been a surge of research on improving the communication efficiency of distributed training. However, little work has been done to systematically understand whether the network is the bottleneck and to what extent. In this paper, we take a first-principles approach to measure and analyze the network performance of distributed training. As expected, our measurement confirms that communication is the component that blocks distributed training from linear scale-out. However, contrary to the common belief, we find that the network is running at low utilization and that if the network can be fully utilized, distributed training can achieve a scaling factor of close to one. Moreover, while many recent proposals on gradient compression advocate over 100x compression ratio, we show that under full network utilization, there is no need for gradient compression in 100 Gbps network. On the other hand, a lower speed network like 10 Gbps requires only 2x-5x gradients compression ratio to achieve almost linear scale-out. Compared to application-level techniques like gradient compression, network-level optimizations do not require changes to applications and do not hurt the performance of trained models. As such, we advocate that the real challenge of distributed training is for the network community to develop high-performance network transport to fully utilize the network capacity and achieve linear scale-out.","PeriodicalId":254313,"journal":{"name":"Proceedings of the Workshop on Network Meets AI & ML","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125150141","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 50