BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160558Z
LOCATION:Track 4
DTSTART;TZID=America/New_York:20201112T174000
DTEND;TZID=America/New_York:20201112T180500
UID:submissions.supercomputing.org_SC20_sess210_ws_indis108@linklings.com
SUMMARY:Computing Bottleneck Structures at Scale for High-Precision Networ
 k Performance Analysis
DESCRIPTION:SCinet, Workshop\n\nComputing Bottleneck Structures at Scale f
 or High-Precision Network Performance Analysis\n\nAmsel, Ros-Giralt, Yella
 mraju, Ezick, von Hofe...\n\nThe Theory of Bottleneck Structures is a rece
 ntly-introduced framework for studying the performance of data networks th
 at describes how local perturbations in one part of the network propagate 
 and interact with one another. This frame- work provides a powerful analyt
 ical tool for network operators to make accurate predictions about network
  behavior and optimize performance. Previous work implemented a software p
 ackage that leveraged the Theory of Bottleneck Structures to address sever
 al network optimization problems, but applied it only to simple examples. 
 In this work, we introduce the first software package capable of scaling t
 he bottleneck structure analysis to production-sized networks. We benchmar
 k our system using logs from ESnet, the Department of Energy high-performa
 nce data network used to connect research institutions in the US. Using th
 e previously published tool as a baseline, we demonstrate that our system 
 achieves vastly improved performance, allowing the bottleneck structure an
 alysis to be applied to rapidly-changing network conditions in real time. 
 We also study the asymptotic performance of our core algorithms, demonstra
 ting strong agreement with theoretical bounds and improvements over algori
 thms of the baseline. These results indicate that the proposed software pa
 ckage is capable of scaling to very large data networks. Overall, we demon
 strate the viability of using bottleneck structures to perform high-precis
 ion bottleneck and flow analysis.\n\nTag: Big Data, Data Analytics, Compre
 ssion, and Management, Datacenter, Networks, Performance/Productivity Meas
 urement and Evaluation, SCinet, Software-defined networking\n\nRegistratio
 n Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR