BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150115Z
LOCATION:D163
DTSTART;TZID=America/Chicago:20181111T113000
DTEND;TZID=America/Chicago:20181111T120000
UID:submissions.supercomputing.org_SC18_sess159_ws_indis104@linklings.com
SUMMARY:Fast Detection of Elephant Flows with Dirichlet-Categorical Infere
 nce
DESCRIPTION:Aditya Gudibanda, Jordi Ros-Giralt, Alan Commike, and Richard 
 Lethin (Reservoir Labs Inc)\n\nThe problem of elephant flow detection is a
  longstanding research area with the goal of quickly identifying flows in 
 a network that are large enough to affect the quality of service of smalle
 r flows. Past work in this field has largely been either domain-specific, 
 based on thresholds for a specific flow size metric, or required several h
 yperparameters, reducing their ease of adaptation to the great variety of 
 traffic distributions present in real-world networks. In this paper, we pr
 esent an approach to elephant flow detection that avoids these limitations
 , utilizing the rigorous framework of Bayesian inference. By observing pac
 kets sampled from the network, we use Dirichlet-Categorical inference to c
 alculate a posterior distribution that explicitly captures our uncertainty
  about the sizes of each flow. We then use this posterior distribution to 
 find the most likely subset of elephant flows under this probabilistic mod
 el. Our algorithm rapidly converges to the optimal sampling rate at a spee
 d O(1/n), where n is the number of packet samples received, and the only h
 yperparameter required is the targeted detection likelihood, defined as th
 e probability of correctly inferring all the elephant flows. Compared to t
 he state-of-the-art based on static sampling rate, we show a reduction in 
 error rate by a factor of 20 times. The proposed method of Dirichlet-Categ
 orical inference provides a novel, powerful framework to elephant flow det
 ection that is both highly accurate and probabilistically meaningful.\n\nT
 ag: Architectures, Networks, Security\n\nRegistration Category: Workshop R
 eg Pass\n\nSession Chairs: Ilya Baldin (Thomas Jefferson National Accelera
 tor Facility); Paola Grosso (University of Amsterdam, Netherlands); Mary H
 ester (Dutch National Institute for Subatomic Physics); and Michelle Zhu (
 Montclair State University)\n\n
END:VEVENT
END:VCALENDAR