{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,31]],"date-time":"2026-03-31T15:55:34Z","timestamp":1774972534924,"version":"3.50.1"},"reference-count":31,"publisher":"Association for Computing Machinery (ACM)","issue":"4","license":[{"start":{"date-parts":[[2026,3,31]],"date-time":"2026-03-31T00:00:00Z","timestamp":1774915200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["Commun. ACM"],"published-print":{"date-parts":[[2026,4,1]]},"abstract":"<jats:p>\n                    Flow optimization is fundamental to computer science, and particularly to computer networks. A prominent example is flow optimization in the global \u201cbackbone networks\u201d that interconnect service providers\u2019 datacenters. We tackle a major gap between theory and practice: While in theoretical models upcoming traffic demands are typically known, in real-world networks such information is rarely available a priori. In practice, addressing this gap often involves\n                    <jats:italic toggle=\"yes\">predicting<\/jats:italic>\n                    upcoming demands and optimizing for these. Using data from production networks, we show that this approach can produce solutions that deviate significantly from the optimum.\n                  <\/jats:p>\n                  <jats:p>\n                    We propose a novel approach: leveraging empirical data to directly\n                    <jats:italic toggle=\"yes\">learn<\/jats:italic>\n                    flow configurations that deliver\n                    <jats:italic toggle=\"yes\">robustly high<\/jats:italic>\n                    performance, bypassing the need for explicit demand prediction. We prove the optimality of our methodology. We further show that by building on recent advances in large-scale optimization and deep learning, our approach enables efficient training on extensive data, picking out intricate patterns in real-world traffic. Through extensive empirical evaluation, we demonstrate that our approach significantly outperforms the state of the art in terms of both solution quality and online runtimes.\n                  <\/jats:p>","DOI":"10.1145\/3765706","type":"journal-article","created":{"date-parts":[[2026,2,18]],"date-time":"2026-02-18T03:47:57Z","timestamp":1771386477000},"page":"80-86","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Learning to Flow (Between Datacenters)"],"prefix":"10.1145","volume":"69","author":[{"ORCID":"https:\/\/orcid.org\/0009-0008-8416-805X","authenticated-orcid":false,"given":"Yarin","family":"Perry","sequence":"first","affiliation":[{"name":"Hebrew University of Jerusalem, Jerusalem, Israel"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9494-6435","authenticated-orcid":false,"given":"Srikanth","family":"Kandula","sequence":"additional","affiliation":[{"name":"Amazon Web Services, Seattle, Washington, United States"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2540-236X","authenticated-orcid":false,"given":"Ishai","family":"Menache","sequence":"additional","affiliation":[{"name":"Microsoft Research, Redmond, Washington, United States"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9336-8351","authenticated-orcid":false,"given":"Michael","family":"Schapira","sequence":"additional","affiliation":[{"name":"Hebrew University of Jerusalem, Jerusalem, Israel"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1972-854X","authenticated-orcid":false,"given":"Aviv","family":"Tamar","sequence":"additional","affiliation":[{"name":"Technion - Israel Institute of Technology, Haifa, Haifa District, Israel"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"320","published-online":{"date-parts":[[2026,3,31]]},"reference":[{"key":"e_1_3_1_2_2","unstructured":"Abilene\/Internet2. (2004); http:\/\/www.internet2.edu\/"},{"key":"e_1_3_1_3_2","unstructured":"Abuzaid F. et al. Contracting wide-area network topologies to solve flow problems quickly. In 18th USENIX Symp. on Networked Systems Design and Implementation. USENIX Association\u00a0(2021) \u00a0175\u2013200; https:\/\/www.usenix.org\/conference\/nsdi21\/presentation\/abuzaid"},{"key":"e_1_3_1_4_2","doi-asserted-by":"crossref","unstructured":"Ahuja S.S. et al. Capacity-efficient and uncertainty-resilient backbone network planning with hose. In Proceedings of the 2021 ACM SIGCOMM 2021 Conf. ACM\u00a0(2021) 547\u2013559.","DOI":"10.1145\/3452296.3472918"},{"key":"e_1_3_1_5_2","doi-asserted-by":"crossref","unstructured":"Applegate D. and Cohen E. Making intra-domain routing robust to changing and uncertain traffic demands: understanding fundamental tradeoffs. In Proceedings of the 2003 Conf. on Applications Technologies Architectures and Protocols for Computer Communications. ACM\u00a0(2003) 313\u2013324.","DOI":"10.1145\/863955.863991"},{"key":"e_1_3_1_6_2","doi-asserted-by":"crossref","unstructured":"Axiotis K. M\u0105dry A. and Vladu A. Faster sparse minimum cost flow by electrical flow localization. In 2021 IEEE 62nd Annual Symp. on Foundations of Computer Science. IEEE (2022) 528\u2013539.","DOI":"10.1109\/FOCS52979.2021.00059"},{"key":"e_1_3_1_7_2","doi-asserted-by":"publisher","DOI":"10.1145\/780542.780599"},{"key":"e_1_3_1_8_2","doi-asserted-by":"crossref","unstructured":"Bogle J. et al. TEAVAR: Striking the right utilization-availability balance in WAN traffic engineering. In Proceedings of the ACM Special Interest Group on Data Communication. ACM\u00a0(2019) 29\u201343.","DOI":"10.1145\/3341302.3342069"},{"key":"e_1_3_1_9_2","volume-title":"Advances in Neural Information Processing Systems,","author":"Hazan E.","year":"2015","unstructured":"Hazan, E., Levy, K., and Shalev-Shwartz, S. Beyond convexity: Stochastic quasi-convex optimization. In Advances in Neural Information Processing Systems,\u00a0Vol. 28. Curran Associates\u00a0(2015); https:\/\/proceedings.neurips.cc\/paper_files\/paper\/2015\/file\/934815ad542a4a7c5e8a2dfa04fea9f5-Paper.pdf"},{"key":"e_1_3_1_10_2","doi-asserted-by":"crossref","unstructured":"Hong C.-Y. et al. Achieving high utilization with software-driven WAN. In Proceedings of the ACM SIGCOMM 2013 Conf. on SIGCOMM. ACM\u00a0(2013) 15\u201326.","DOI":"10.1145\/2486001.2486012"},{"key":"e_1_3_1_11_2","doi-asserted-by":"crossref","unstructured":"Hong C.-Y. et al. B4 and after: Managing hierarchy partitioning and asymmetry for availability and scale in google\u2019s software-defined WAN. In Proceedings of the 2018 Conf. of the ACM Special Interest Group on Data Communication. ACM\u00a0(2018) 74\u201387.","DOI":"10.1145\/3230543.3230545"},{"key":"e_1_3_1_12_2","doi-asserted-by":"crossref","unstructured":"Jain S. et al. B4: Experience with a globally deployed software defined WAN. In Proceedings of the ACM SIGCOMM 2013 Conf. on SIGCOMM. ACM\u00a0(2013) 3\u201314.","DOI":"10.1145\/2486001.2486019"},{"key":"e_1_3_1_13_2","doi-asserted-by":"crossref","unstructured":"Kandula S. Menache I. Schwartz R. and Babbula S.R. Calendaring for wide area networks. In Proceedings of the 2014 ACM Conf. on SIGCOMM. ACM\u00a0(2014) 515\u2013526.","DOI":"10.1145\/2619239.2626336"},{"key":"e_1_3_1_14_2","doi-asserted-by":"publisher","DOI":"10.1145\/1328911.1328924"},{"key":"e_1_3_1_15_2","doi-asserted-by":"crossref","unstructured":"Kelner J.A. Miller G.L. and Peng R. Faster approximate multicommodity flowusing quadratically coupled flows. In Proceedings of the Forty-Fourth Annual ACM Symp. on Theory of Computing. ACM\u00a0(2012) 1\u201318.","DOI":"10.1145\/2213977.2213979"},{"key":"e_1_3_1_16_2","unstructured":"Kingma D.P. and Ba J. Adam: A method for stochastic optimization. arXiv:1412.6980 [cs.LG] (2014); https:\/\/arxiv.org\/abs\/1412.6980"},{"key":"e_1_3_1_17_2","doi-asserted-by":"publisher","DOI":"10.1109\/JSAC.2011.111002"},{"key":"e_1_3_1_18_2","doi-asserted-by":"publisher","DOI":"10.1080\/1055678031000111236"},{"key":"e_1_3_1_19_2","unstructured":"Krishnaswamy U. et al. OneWAN is better than two: Unifying a split WAN architecture. In 20th USENIX Symp. on Networked Systems Design and Implementation. USENIX Association\u00a0(2023) 515\u2013529; https:\/\/www.usenix.org\/conference\/nsdi23\/presentation\/krishnaswamy"},{"key":"e_1_3_1_20_2","unstructured":"Kumar P. et al. Semi-oblivious traffic engineering: The road not taken. In 15th USENIX Symp. on Networked Systems Design and Implementation. USENIX Association\u00a0(2018) 157\u2013170; https:\/\/www.usenix.org\/conference\/nsdi18\/presentation\/kumar"},{"key":"e_1_3_1_21_2","doi-asserted-by":"crossref","unstructured":"Lan T. Kao D. Chiang M. and Sabharwal A. An axiomatic theory of fairness in network resource allocation. In 2010 Proceedings IEEE INFOCOM (2010) 1\u20139.","DOI":"10.1109\/INFCOM.2010.5461911"},{"key":"e_1_3_1_22_2","doi-asserted-by":"publisher","DOI":"10.1006\/jcss.1995.1020"},{"key":"e_1_3_1_23_2","doi-asserted-by":"publisher","DOI":"10.1145\/331524.331526"},{"key":"e_1_3_1_24_2","doi-asserted-by":"crossref","unstructured":"Liu H.H. et al. Traffic engineering with forward fault correction. In Proceedings of the 2014 ACM Conf. on SIGCOMM.\u00a0ACM\u00a0(2014) 527\u2013538.","DOI":"10.1145\/2619239.2626314"},{"key":"e_1_3_1_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/90.879343"},{"key":"e_1_3_1_26_2","doi-asserted-by":"crossref","unstructured":"Narayanan D. et al. Solving large-scale granular resource allocation problems efficiently with POP. In Proceedings of the ACM SIGOPS 28th Symp. on Operating Systems Principles. ACM\u00a0(2021) 521\u2013537.","DOI":"10.1145\/3477132.3483588"},{"issue":"3","key":"e_1_3_1_27_2","first-page":"519","article-title":"Minimization methods for nonsmooth convex and quasiconvex functions","volume":"29","author":"Nesterov Y.E.","year":"1984","unstructured":"Nesterov, Y.E. Minimization methods for nonsmooth convex and quasiconvex functions. Matekon 29, 3 (1984), 519\u2013531.","journal-title":"Matekon"},{"key":"e_1_3_1_28_2","unstructured":"Perry Y. et al. DOTE: Rethinking (predictive)WAN traffic engineering. In 20th USENIX Symp. on Networked Systems Design and Implementation. USENIX Association\u00a0(2023) 1557\u20131581; https:\/\/www.usenix.org\/conference\/nsdi23\/presentation\/perry"},{"key":"e_1_3_1_29_2","doi-asserted-by":"crossref","unstructured":"Racke H. Minimizing congestion in general networks. In The 43rd Annual IEEE Symp. on Foundations of Computer Science 2002. Proceedings. IEEE (2002) 43\u201352.","DOI":"10.1109\/SFCS.2002.1181881"},{"key":"e_1_3_1_30_2","doi-asserted-by":"publisher","DOI":"10.1137\/1.9781611976595"},{"key":"e_1_3_1_31_2","doi-asserted-by":"publisher","DOI":"10.1145\/1111322.1111341"},{"key":"e_1_3_1_32_2","doi-asserted-by":"publisher","DOI":"10.1109\/TNET.2016.2594235"}],"container-title":["Communications of the ACM"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3765706","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,31]],"date-time":"2026-03-31T13:55:25Z","timestamp":1774965325000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3765706"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,31]]},"references-count":31,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2026,4,1]]}},"alternative-id":["10.1145\/3765706"],"URL":"https:\/\/doi.org\/10.1145\/3765706","relation":{},"ISSN":["0001-0782","1557-7317"],"issn-type":[{"value":"0001-0782","type":"print"},{"value":"1557-7317","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,31]]},"assertion":[{"value":"2026-03-31","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}