[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120238-en":3,"doc-seo-120238-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},120238,7971461740886,"Theodore","https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2",8,"Research & Report","L3DML - Facilitating Geo-Distributed Machine Learning in Network Layer","Geo-Distributed Machine Learning (GDML) trains large-scale models across geographically dispersed datacenters, but WAN bandwidth limits gradient aggregation and the straggler problem degrades progress. Existing GDML approaches show conflicting behavior, and in-network computing is often confined to single-datacenter settings. L3DML leverages P4-based SDN to enable location-specific in-network gradient aggregation without parameter servers, integrates lossless gradient transmission in switches, and mitigates stragglers via a DRL model with rate-synchronization routing. Experiments on Intel Tofino switches and a Spirent emulator show improved goodput, accuracy, and training speed for large-scale GDML.","Missouri University of Science and Technology  \nScholars' Mine  \n\n| Computer Science Faculty Research & Creative Works | Computer Science |\n| --- | --- |\n| 01 Jan 2024\u003Cbr>L3DML: Facilitating Geo-Distributed Machine Learning in Network Layer\u003Cbr>Xindi Hou\u003Cbr>Shuai Gao Ningchun Liu Fangtao Yao\u003Cbr>[et. al. For a complete list of authors](et. al. For a complete list of authors), see [https://](https://)scholarsmine. mst. edu/comsci_facwork/1950\u003Cbr>Follow this and additional works at: [https://scholarsmine.mst.edu/comsci_facwork](https://scholarsmine.mst.edu/comsci_facwork)\u003Cbr> Part of the Computer Sciences Commons |  |\n\nRecommended Citation  \nX. Hou et al., \"L3DML: Facilitating Geo-Distributed Machine Learning in Network Layer,\" IEEE Transactions on Network and Service Management, Institute of Electrical and Electronics Engineers, Jan 2024. The definitive version is available at [https://doi.org/10.1109/TNSM.2024.3509031](https://doi.org/10.1109/TNSM.2024.3509031)  \n[This Article-Journal is brought to you for free and open access by Scholars](This Article-Journal is brought to you for free and open access by Scholars)' Mine. It has been accepted for inclusion in Computer Science Faculty Research & Creative Works by an authorized administrator of Scholars'Mine. This work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the permission of the copyright holder. For more information, please contact [scholarsmine@mst.edu](scholarsmine@mst.edu).  \nThis article has been accepted for publication in IEEE Transactions on Network and Service Management. This is the author's version which has not been fully edited and content may change prior to final publication. Citation information: DOI 10. 1109/TNSM.2024.3509031  \nIEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT 1  \nL3DML: Facilitating Geo-Distributed Machine Learning in Network Layer  \nXindi Hou, Shuai Gao∗ , Member, IEEE, Ningchun Liu, Fangtao Yao, Bo Lei, Hongke Zhang, Fellow, IEEE,  \nand Sajal K. Das, Fellow, IEEE  \nAbstract—Geo-Distributed Machine Learning (GDML) aims to train large-scale machine learning models across geographically dispersed datacenters. However, the performance of GDML systems is constrained by the limited Wide Area Network (WAN) bandwidth and the presence of the straggler problem. Existing GDML designs often show contradictory effects in addressing these challenges, while in-network computing attempts are typically restricted to single datacenter environments rather than the more complex GDML scenarios. To overcome these limitations, this paper proposes L3DML to facilitate GDML using the P4-based Software-defined Network (SDN). Our approach incorporates three key innovations. Firstly, we introduce a novel network addressing scheme that enables location-specific innetwork gradient aggregation for GDML, eliminating the need for parameter servers. Secondly, we utilize the P4 data plane to integrate lossless gradient transmission within switches. Thirdly, we address the straggler problem by employing a unique Deep Reinforcement Learning (DRL) model set and a corresponding rate synchronization routing approach. L3DML is implemented on a prototype system consisting of several Intel Tofino switches and the Spirent network emulator. Experimental results indicate that L3DML outperforms existing solutions in terms of goodput, model accuracy, and training speed gain for large-scale GDML.  \nIndex Terms—Distributed training, Network-layer addressing, In-network computing, P4, Deep reinforcement learning  \nI. INTRODUCTION  \nAI-assisted applications leverage advanced AI models and  \nlarge volumes of data to achieve enhanced performance. To accelerate model training, machine learning practitioners employ distributed training (DT), enabling training jobs to be distributed among multiple workers. This significantly reduces the training time to just a few hours or days [1] . However, the raw data is geographically dispersed a","cbCaieSadYNezJhC","https://ap.wps.com/l/cbCaieSadYNezJhC","pdf",4074904,1,19,"English","en",105,"# Abstract\n# Index Terms\n# I. Introduction\n## Geo-Distributed Machine Learning overview\n## Performance challenges: WAN bandwidth and straggler problem\n## Paper contributions and motivation","[{\"question\":\"What problem does L3DML address in geo-distributed machine learning?\",\"answer\":\"It addresses GDML performance limits caused by limited WAN bandwidth during gradient aggregation and by the straggler problem that slows distributed training progress.\"},{\"question\":\"How does L3DML reduce the need for parameter servers?\",\"answer\":\"It introduces a location-specific network addressing scheme that supports in-network gradient aggregation, eliminating the parameter-server requirement.\"},{\"question\":\"How does L3DML handle the straggler problem?\",\"answer\":\"It uses a deep reinforcement learning (DRL) model combined with a rate synchronization routing approach to improve training under straggler conditions.\"}]","L3DML - Facilitating Geo-Distributed Machine Learning in Network Layer | PDF",1785728931,48,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"l3dml-facilitating-geo-distributed-machine-learning-in-network-layer","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/l3dml-facilitating-geo-distributed-machine-learning-in-network-layer/120238/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does L3DML address in geo-distributed machine learning?","Question",{"text":75,"@type":76},"It addresses GDML performance limits caused by limited WAN bandwidth during gradient aggregation and by the straggler problem that slows distributed training progress.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does L3DML reduce the need for parameter servers?",{"text":80,"@type":76},"It introduces a location-specific network addressing scheme that supports in-network gradient aggregation, eliminating the parameter-server requirement.",{"name":82,"@type":73,"acceptedAnswer":83},"How does L3DML handle the straggler problem?",{"text":84,"@type":76},"It uses a deep reinforcement learning (DRL) model combined with a rate synchronization routing approach to improve training under straggler conditions.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},"General","general"]