<html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"></head><body><div>Ciao Dave (&all)</div><div><br></div><div>While I'd love to do a bit more on the topic, I recently joined Huawei, and while I still work on networking, this one is not so aligned with my current set of activities... </div><div><br></div><div>so I cannot unfortunately help moving forward :/</div><div><br></div><div>Best,</div><div>D.</div><div><br></div><div>PS fyi Also I tend to read this email less freqiently than @huawei</div><div><br></div><div id="composer_signature"><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"><div style="font-size:85%;color:#575757">Sent from my new-but-still-not-so-smart-phone. Excuse my typos and its random fixes</div><div style="font-size:85%;color:#575757"><br></div><div style="font-size:85%;color:#575757">Oo Chair holder NewNet@Paris</div><div style="font-size:85%;color:#575757"> > Professor, Telecom ParisTech</div><div style="font-size:85%;color:#575757">~ Professor, Ecole Polytechnique </div><div><br></div></div><div><br></div><div style="font-size:100%;color:#000000"></div><div style="font-size:100%;color:#000000"><!-- originalMessage --><div>-------- Original message --------</div><div>From: Dave Taht <dave.taht@gmail.com> </div><div>Date: 12/6/18 04:18 (GMT+01:00) </div><div>To: bloat <bloat@lists.bufferbloat.net>, Cake List <cake@lists.bufferbloat.net>, aqm@ietf.org </div><div>Subject: paper: per flow fairness in a data center network </div><div><br></div></div>While I strongly agree with their premise:<br><br>"Multi-tenant DCNs cannot rely on specialized protocols and mechanisms<br>that assume single ownership and end-system compliance. It is<br>necessary rather to implement general, well-understood mechanisms<br>provided as a network service that require as few assumptions about DC<br>workload as possible."<br><br>... And there's a solid set of links to current work, and a very<br>interesting comparison to pfabric, their DCTCP emulation is too flawed<br>to be convincing, and we really should get around to making the ns2<br>fq_codel emulation fully match reality. This is also a scenario where<br>I'd like to see cake tried, to demonstrate the effectiveness (or not!)<br>of 8 way set associative queuing, cobalt, per host/per flow fq, etc,<br>vs some of the workloads they outline.<br><br>https://perso.telecom-paristech.fr/drossi/paper/rossi18hpsr.pdf<br><br>-- <br><br>Dave Täht<br>CTO, TekLibre, LLC<br>http://www.teklibre.com<br>Tel: 1-831-205-9740<br></body></html>