BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20170326T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20161030T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260428T225250Z
UID:58175efb240f0682917052@ist.ac.at
DTSTART:20161221T134500
DTEND:20161221T144500
DESCRIPTION:Speaker: Malte Schwarzkopf\nhosted by Tom Henzinger\nAbstract: 
 Scheduling tasks on clusters in large-scale datacenters is challenging:\nt
 housands of tasks must be cleverly and quickly placed to achieve good appl
 ication-level performance and high resource utilization. Centralized datac
 enter schedulers can make high-quality placement decisions\, but today the
 se high-quality placements come at the cost of high latency at scale\, whi
 ch degrades response time for interactive tasks and reduces resource utili
 zation.\n\nIn this talk I present Firmament\, a centralized scheduler that
  scales to over ten thousand machines\, even though it performs a computat
 ionally expensive min-cost max-flow (MCMF) optimization that continuously 
 reschedules all tasks. To achieve this\, Firmament automatically chooses b
 etween different MCMF algorithms\, solves the optimization problem increme
 ntally when possible\, and applies problem-specific optimizations.\n\nExpe
 riments with a Google workload trace from a 12\,500-machine cluster show t
 hat Firmament places tasks in hundreds of milliseconds\, and that Firmamen
 t improves placement latency by 20x over Quincy\, a prior centralized sche
 duler using the same MCMF optimization. Moreover\, even though Firmament i
 s centralized\, it matches the placement latency of distributed schedulers
  for workloads of short tasks. Finally\, Firmament exceeds the placement q
 uality of four widely-used centralized and distributed schedulers on a rea
 l-world cluster\, improving batch task response time by 6x.\n\nReference:\
 nI. Gog\, M. Schwarzkopf\, A. Gleave\, R. N. M. Watson\, S. Hand. "Firmame
 nt:\nFast\, Centralized Cluster Scheduling at Scale". In Proceedings of th
 e 12th USENIX Symposium on Operating Systems Design and Implementation (OS
 DI)\, Savannah\, GA\, USA\, November 2016\, p. 99--115.
LOCATION:Mondi Seminar Room 2\, Central Building\, ISTA
ORGANIZER:abonvent@ist.ac.at
SUMMARY:Malte Schwarzkopf: Firmament: Fast\, Centralized Cluster Scheduling
  at Scale
URL:https://talks-calendar.ista.ac.at/events/240
END:VEVENT
END:VCALENDAR