BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:2e8bfbddab5738952bea04ea9b20c878
DTSTAMP:20260408T060234Z
SUMMARY:Mercury: Hybrid Centralized and Distributed Scheduling in Large Sha
 red Clusters (K. Karanasos )
DESCRIPTION:Abstract\nDatacenter-scale computing for analytics workloads is
  increasingly  common. High operational costs force heterogeneous applicat
 ions to share  cluster resources for achieving economy of scale. Scheduli
 ng such large  and diverse workloads is inherently hard\, and existing app
 roaches  tackle this in two alternative ways: 1) centralized solutions of
 fer  strict\, secure enforcement of scheduling invariants (e.g.\, fairness
 \,  capacity) for heterogeneous applications\, 2) distributed solutions of
 fer  scalable\, efficient scheduling for homogeneous applications. We  a
 rgue that these solutions are complementary\, and advocate a blended  appr
 oach. Concretely\, we propose Mercury\, a hybrid resource management  fram
 ework that supports the full spectrum of scheduling\, from  centralized t
 o distributed. Mercury exposes a programmatic interface  that allows appli
 cations to trade-off between scheduling overhead and  execution guarantee
 s. Our framework harnesses this flexibility by  opportunistically utilizin
 g resources to improve task  throughput. Experimental results on producti
 on-derived workloads show  gains of over 35% in task throughput. These ben
 efits can be translated  by appropriate application and framework policie
 s into job throughput or  job latency improvements. We have implemented an
 d are currently  contributing Mercury as an extension of Apache Hadoop / 
 YARN. This  work will appear in USENIX ATC and is a joint work with Srira
 m Rao\,  Carlo Curino\, Chris Douglas\, Kishore Chaliparambil\, Giovanni F
 umarola\,  Solom Heddaya\, Raghu Ramakrishnan\, and Sarvesh Sakalanaga.\n
 Bio\nKonstantinos Karanasos joined the Cloud and Information Services  Lab
  (CISL) at Microsoft as a Senior Scientist in March 2014\, where he is  wo
 rking on cloud-scale cluster resource management\, distributed data  plat
 forms\, and query processing and optimization. He is currently based  in C
 ambridge\, UK\, where he is also collaborating with the Systems &  Networ
 king group of Microsoft Research. Prior to joining Microsoft\,  Konstantin
 os was a postdoctoral researcher at IBM Almaden Research  Center\, where 
 he was member of the Big Data analytics group. At Almaden\,  he was workin
 g on Jaql\, a platform for analyzing large datasets in  parallel using Had
 oop’s MapReduce framework. He built a system that  extended Jaql by add
 ing dynamic optimization capabilities to it. Part of  his work at Almaden 
 was transferred to IBM BigInsights product.  Konstantinos obtained his Ph
 D in Computer Science from Inria\, France\,  working on view-based techniq
 ues for semi-structured data\, under the  supervision of Ioana Manolescu 
 and Francois Goasdoue. Prior to that\, he  received his Diploma in Electri
 cal and Computer Engineering from the  National Technical University of At
 hens\, Greece\, where he completed his  Diploma thesis under the supervis
 ion of Timos Sellis.
URL:https://www.imperial.ac.uk/events/104161/mercury-hybrid-centralized-and
 -distributed-scheduling-in-large-shared-clusters-k-karanasos/
DTSTART;TZID=Europe/London:20150730T140000
DTEND;TZID=Europe/London:20150730T150000
LOCATION:United Kingdom
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
DTSTART:20150730T140000
TZNAME:BST
TZOFFSETTO:+0100
TZOFFSETFROM:+0100
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR