BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium: Building Better Datacenters - The Quest for Low Lat
 ency
DTSTART:20210412T160000
DTEND:20210412T170000
DTSTAMP:20260510T164834Z
UID:75ca228eb692068b0234101e93fcfb7b607ef38caa59dab89031ea58
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Simon Peter - University of Texas at Austin\n\nAbstract\nA
 s datacenter applications grow in number and complexity\, datacenter-inter
 nal service latency requirements are dropping into the microsecond range. 
 Providing consistent microsecond-scale service latencies at increasing dat
 acenter utilization is difficult\, especially at scale\, where failures ar
 e common. Operating system functionality on the service critical path ofte
 n incurs high\, millisecond-scale overhead\, and introduces even longer qu
 eueing delay\nas utilization increases and during fail-over. My research a
 ims to dramatically lower service latencies under rising utilization by co
 -designing hardware and operating system functionality to remove these ove
 rheads from the critical path\, even when failures are common.\n\nMy recen
 t focus has been on building low latency and available storage systems. Th
 e adoption of low latency persistent memory modules (PMMs)\nin datacenter 
 servers upends the long-established model of remote storage for distribute
 d file systems. Instead\, by colocating computation with PMM storage we ca
 n provide applications with much lower IO and application failover latenci
 es\, while offering strong consistency. I present Assise\, a new distribut
 ed file system\, based on a persistent\, replicated coherence protocol tha
 t manages client-local PMM as a linearizable and crash-recoverable cache b
 etween applications and slower (and possibly remote) storage. Assise maxim
 izes locality for all file IO by carrying out IO on process-local\, socket
 -local\, and\nclient-local PMM whenever possible. Assise minimizes coheren
 ce overhead by maintaining consistency at IO operation granularity\, rathe
 r than at fixed block sizes. Assise improves IO latency\, throughput\, and
  fail-over time by an order of magnitude versus the state-of-the-art\, whi
 le providing stronger consistency semantics. I finish with an overview of 
 further research in this space\, and an outlook to impending energy constr
 aints of large scale systems\, leading to a future research agenda in ener
 gy-resilient system design.\n\nBio\nSimon is an assistant professor in com
 puter science at The University of Texas at Austin. Simon works to dramati
 cally improve data center efficiency and reliability by designing\, buildi
 ng\, and evaluating new alternatives for their hardware and software compo
 nents. Simon currently co-designs networking and storage stacks with new h
 ardware technologies to reduce service latencies by orders of magnitude be
 yond today's capabilities.\n\nSimon is the director of the Texas Systems R
 esearch Consortium\, where he collaborates closely with industry to shape 
 the future of cloud computing. Simon's work is supported by VMware\, Micro
 soft Research\, Huawei\, Google\, Citadel Securities\, and Arm. Simon rece
 ived the SIGOPS Hall of Fame award in 2020. He was twice awarded the Jay L
 epreau Best Paper Award\, in 2014 and 2016\, an IEEE Micro Top Pick Honora
 ble Mention in 2021\, and a Memorable Paper Award in 2018. He received an 
 NSF CAREER Award and he is a Sloan research fellow. Before joining UT Aust
 in in 2016\, Simon was a research associate at the University of Washingto
 n from 2012-2016. He received a Ph.D. in Computer Science from ETH Zurich 
 in 2012.\n\nMore information
LOCATION:https://epfl.zoom.us/j/88487368284?pwd=TU5QNGdqSHVnTFREaFgrelFaTU
 pIdz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR