BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/525
DTSTAMP:20230914T125928Z
SUMMARY:Randomized Load Balancing in Large Bandwidth Sharing Systems
DESCRIPTION:Speaker: Ravi R. Mazumdar (University of Waterloo\nDepartment o
 f Electrical and\nComputer Engineering\nWaterloo\, ON N2L 3G1\nCanada)\n\n
 Abstract: \nAbstract: Processor sharing models occur in a wide variety of 
 situations for example in models of internet bottlenecks. They are good mo
 dels for bandwidth sharing as well as being solutions to NUM for logarithm
 ic utilities. In addition they possess the desirable stochastic property o
 f the stationary distribution being insensitive to the service time distri
 bution. In this talk I will discuss new advances in understanding and char
 acterizing the behavior of randomized routing to PS servers that are heter
 ogeneous in terms of their server speeds.\n\nWe will rst discuss the so-ca
 lled Power-of-two rule in the homogeneous case of identical servers where 
 routing to the least occupied server amongst two randomly chosen servers r
 esults in a very low server occupancy and a so-called propagation of chaos
  or asymptotic independence. There were no known characterizations about t
 he heterogeneous case. In the heterogeneous case we will see that the stab
 ility region for randomized routing is strictly included in the maximal st
 ability region that can be achieved by state independent routing. Therefor
 e the average sojourn of tasks can be longer in randomized routing in hete
 rogeneous systems. When the system is stable we completely characterize th
 e steady-state behavior of the server occupancies and show that it exhibit
 s super-exponential decay and asymptotic independence among servers. To ov
 ercome the reduction in the stability region we show that a combination of
  state independent routing (biased sampling) to a server class combined wi
 th JSQ within the class recovers the stability region as well as the benet
 s of small server occupancies.\n\nThe techniques are based on a mean eld a
 nalysis and an ansatz based on propagation of chaos that then establishes 
 the asymptotic independence between servers (joint work with Arpan Mukhopa
 dhyay (Waterloo).\n\nBiography: The speaker was educated at the Indian Ins
 titute of Technology\, Bombay (B.Tech\, 1977)\, Imperial College\, London 
 (MSc\, DIC\, 1978) and obtained his PhD under A. V. Balakrishnan at UCLA i
 n 1983. He is currently a University Research Chair Professor in the Dept.
  of ECE at the University of Waterloo\, Ont.\, Canada where he has been si
 nce September 2004. Prior to this he was Professor of ECE at Purdue Univer
 sity\, West Lafayette\, USA. Since 2012 he is a D.J. Gandhi Distinguished 
 Visiting Professor at the Indian Institute of Technology\, Bombay\, India.
  He is a Fellow of the IEEE and the Royal Statistical Society. He is a rec
 ipient of the INFOCOM 2006 Best Paper Award and was runner-up for the Best
  Paper Award at INFOCOM 1998. His research interests are in modeling\, con
 trol\, and performance analysis of both wireline and wireless networks\, a
 nd in applied probability and stochastic analysis with applications to que
 ueing\, ltering\, and optimization.\n\n \n
URL:https://www.tcs.tifr.res.in/web/events/525
DTSTART;TZID=Asia/Kolkata:20140807T160000
DTEND;TZID=Asia/Kolkata:20140807T170000
LOCATION:AG-80
END:VEVENT
END:VCALENDAR