BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163603Z
LOCATION:501
DTSTART;TZID=America/Denver:20191117T165500
DTEND;TZID=America/Denver:20191117T172000
UID:submissions.supercomputing.org_SC19_sess103_ws_indis114@linklings.com
SUMMARY:Sample Transfer Optimization with Adaptive Deep Neural Network
DESCRIPTION:Workshop\n\nSample Transfer Optimization with Adaptive Deep Ne
 ural Network\n\nSapkota, Arifuzzaman, Arslan\n\nTransfer configurations pl
 ay a crucial role in achieving desirable performance in high-speed network
 s where suboptimal settings could lead to poor transfer throughput. Howeve
 r, discovering the optimal configuration for a given transfer task is a di
 fficult problem as it depends on various factors including dataset charact
 eristics and network settings. The state-of-the-art transfer tuning soluti
 ons rely on sample transfers and evaluate different transfer configuration
 s in attempt to discover the optimal one in real-time. Yet, current approa
 ches to run sample transfers incur significant delay and measurement error
 s, thus limit the gain offered by tuning algorithms. In this paper, we tak
 e advantage of feed forward deep neural network (DNN) to minimize executio
 n time of sample transfers without sacrificing measurement accuracy. To ac
 hieve this goal, we collected 115K data transfer logs in four networks and
  trained multiple DNNs that can predict convergence time of transfers by a
 nalyzing real-time throughput metrics. The results gathered in various net
 works with rich set of transfer configurations indicate that DNN can reduc
 e error rate by up to 50% compared to the state-of-the-art solution while 
 achieving similar sample transfer execution time in most cases. Moreover, 
 by tuning its hyperparameters and model settings, one can achieve low exec
 ution time and error rate based on the specific needs of the user or appli
 cation.\n\nTag: Workshop Reg Pass, Big Data, Data Analytics, Datacenter, N
 etworks, Software-defined networking\n\nRegistration Category: Workshop Re
 g Pass, Big Data, Data Analytics, Datacenter, Networks, Software-defined n
 etworking
URL:https://sc19.supercomputing.org/presentation/?id=ws_indis114&sess=sess
 103
END:VEVENT
END:VCALENDAR