BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Case for Decentralized Scheduling in Modern Datacenters - Smit
 a Vijayakumar\, Systems Research Group\, Cambridge University Computer Lab
 oratory
DTSTART:20250501T140000Z
DTEND:20250501T150000Z
UID:TALK228694@talks.cam.ac.uk
CONTACT:Richard Mortier
DESCRIPTION:"Join on MS Teams":https://teams.microsoft.com/l/meetup-join/1
 9%3ameeting_NGI3NThhZGMtNDFlNS00ZTJhLWJlYWUtYzAyYWIzZGMwODY4%40thread.v2/0
 ?context=%7b%22Tid%22%3a%2249a50445-bdfa-4b79-ade3-547b4f3986e9%22%2c%22Oi
 d%22%3a%22c74ff4ca-98fe-4b28-9889-e119acc12f30%22%7d\n\nThe growing demand
  for data centre resources and the slower evolution of their hardware have
  led to clusters operating at high utilisation. In this talk\, I will exam
 ine how current schedulers perform under such conditions. I will discuss h
 ow centralised schedulers struggle to scale under high load due to the sig
 nificant network traffic caused by continuously transferring up-to-date no
 de data. Conversely\, distributed schedulers scale well but lack a global 
 cluster view\, leading to suboptimal task allocations. As a result\, exist
 ing schedulers impose up to three times longer wait times on tail tasks\, 
 which increases job completion times.\n\nI will then introduce our work on
  decentralised scheduling\, focusing on performance\, scalability\, and lo
 ad balancing. These schedulers have been largely under-explored due to the
 ir design complexity. However\, we demonstrate that Murmuration\, our job-
 aware decentralised scheduler\, achieves high performance under both norma
 l and high load despite its simple approach using approximate load informa
 tion. It reduces communication overhead between nodes and schedulers while
  still achieving balanced cluster load distribution. By the end of this ta
 lk\, I hope to convince you that decentralised schedulers with approximate
  knowledge strike the right balance between performance and scalability\, 
 making them a practical solution for today’s highly utilised data centre
 s. \n\nBio: Smita Vijayakumar recently completed her PhD from the Departme
 nt of Computer Science and Technology at the University of Cambridge\, und
 er the supervision of Evangelia Kalyvianaki. As a part of her thesis\, she
  developed a novel decentralised scheduling framework to reduce tail task 
 latencies in highly utilised data centres. She has over twelve years of in
 dustry experience working on networking\, cloud computing\, and distribute
 d systems. She also has an MS from The Ohio State University\, where her w
 ork investigated cloud resource allocation to bottleneck stages for proces
 sing streaming applications. Her research has been published in top-tier c
 onferences\, and also as a book. She has also been actively involved in me
 ntoring\, teaching\, and community leadership\, including founding Women W
 ho Go\, India.
LOCATION:Computer Lab\, FW11 and Online (MS Teams link below)
END:VEVENT
END:VCALENDAR