BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Load Balancing under Data Locality: Extending Mean-Field Framework
  to Constrained Large-Scale Systems - Dr Debankur Mukherjee\, Georgia Tech
DTSTART:20231004T130000Z
DTEND:20231004T140000Z
UID:TALK204667@talks.cam.ac.uk
CONTACT:Dr Varun Jog
DESCRIPTION:Large-scale parallel-processing infrastructures such as data c
 enters and cloud networks form the cornerstone of the modern digital envir
 onment. Central to their efficiency are resource management policies\, esp
 ecially load balancing algorithms (LBAs)\, which are crucial for meeting s
 tringent delay requirements of tasks. A contemporary challenge in designin
 g LBAs for today's data centers is navigating data locality constraints th
 at dictate which tasks are assigned to which servers. These constraints ca
 n be naturally modeled as a bipartite graph between servers and various ta
 sk types. Most LBA heuristics lean on the mean-field approximation's accur
 acy. However\, the non-exchangeability among servers induced by the data l
 ocality invalidates this mean-field framework\, causing real-world system 
 behaviors to significantly diverge from theoretical predictions. From a fo
 undational standpoint\, advancing our understanding in this domain demands
  the study of stochastic processes on large graphs\, thus needing fundamen
 tal advancements in classical analytical tools. \n\n \n\nIn this presentat
 ion\, we will delve into recent advancements made in extending the accurac
 y of mean-field approximation for a broad class of graphs. In particular\,
  we will talk about how to design resource-efficient\, asymptotically opti
 mal data locality constraints and how the system behavior changes fundamen
 tally\, depending on whether the above bipartite graph is an expander\, a 
 spatial graph\, or is inhomogeneous in nature.\n\n \n\nBio:\n\nDebankur Mu
 kherjee is the Leo and Louise Benatar Early Career Professor and Assistant
  Professor in the H. Milton Stewart School of Industrial and Systems Engin
 eering at the Georgia Institute of Technology. Before joining Georgia Tech
  in 2019\, he was a Prager assistant professor for a year in the Division 
 of Applied Mathematics at Brown University. Debankur got his Ph.D. in Stoc
 hastic Operations Research from the Eindhoven University of Technology in 
 the Netherlands. Debankur’s research spans the area of applied probabili
 ty\, at the interface of stochastic processes and computer science\, with 
 applications to performance analysis\, online algorithms\, and machine lea
 rning. His primary focus is to develop a foundational understanding of the
  challenges that arise in large-scale systems\, such as data centers and c
 loud networks. His work was a finalist in the INFORMS JFIG paper competiti
 on in 2022 and INFORMS George Nicholson Student Paper Competition 2023 and
  received the Best Paper Award at ACM SIGMETRICS 2023 and the Best Student
  Paper Award at ACM SIGMETRICS 2018. His research has been funded by the N
 SF and he is currently serving on the editorial boards of Stochastic Syste
 ms\, QUESTA\, and Stochastic Models.
LOCATION:MR5\, CMS Pavilion A
END:VEVENT
END:VCALENDAR