BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160023Z
LOCATION:Poster Module
DTSTART;TZID=America/New_York:20201117T083000
DTEND;TZID=America/New_York:20201117T170000
UID:submissions.supercomputing.org_SC20_sess342_drs106@linklings.com
SUMMARY:The Coming of Age of Multithreaded High-Performance Communication
DESCRIPTION:Doctoral Showcase, Posters\n\nThe Coming of Age of Multithread
 ed High-Performance Communication\n\nZambre, Chandramowlishwaran\n\nThe su
 percomputing community holds an outdated view: the network is a single dev
 ice. Modern interconnects, however, feature multiple network hardware cont
 exts that serve as parallel interfaces into the network from a single node
 . Additionally, as we are approaching the limits of a single network link&
 rsquo;s throughput, supercomputers are deploying multiple NICs per node to
  accommodate for higher bandwidth per node. Hence, the modern reality is t
 hat the network features lots of parallelism. The outdated view drasticall
 y hurts the communication performance of the MPI+threads model, which is b
 eing increasingly adopted over the traditional MPI-everywhere model to bet
 ter map to modern processors that feature a lesser share of resource per c
 ore than previous processors. Domain scientists typically do not expose lo
 gical parallelism in their MPI+threads communication, and MPI libraries st
 ill use conservative approaches, such as a global critical section, to mai
 ntain MPI’s ordering constraints, thus serializing access to the par
 allel network resources and limiting performance. The goal of this dissert
 ation is to dissolve the communication bottleneck in MPI+threads. Existing
  solutions either sacrifice correctness for performance or jump to MPI sta
 ndard extensions without fairly comparing the capabilities of the existing
  standard. The holistic bottom-up analyses in this dissertation first inve
 stigates the limits of multithreaded communication on modern network hardw
 are, then devises a new MPI-3.1 implementation with virtual communication 
 interfaces (VCIs) for fast MPI+threads communication. The domain scientist
  can use the VCIs either explicitly (MPI Endpoints) or implicitly (MPI-3.1
 ). The dissertation compares the two solutions through both performance an
 d usability lenses.\n\nRegistration Category: Tech Program Reg Pass, Exhib
 its Reg Pass
END:VEVENT
END:VCALENDAR