BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Science Data Processor for the Square Kilometre Array Telescope - 
 Bojan Nikolic - University of Cambridge MRAO
DTSTART:20141022T130000Z
DTEND:20141022T140000Z
UID:TALK54963@talks.cam.ac.uk
CONTACT:David Greaves
DESCRIPTION:Square Kilometre Array (SKA) is planned to be\, by a large fac
 tor\, the\nlargest and most sensitive radio telescope ever constructed. In
  large\npart it has been made possible by advances in electronics and comp
 uting\nallowing simpler and smaller mechanical structures to record the ra
 dio\nsignals in exchange for greater demands on the computing to reconstru
 ct\nthe images.\n\nThe Science Data Processor (SDP) is the part of the SKA
  that will take\nthe signal stream of around 10 TeraByte/s from the digita
 l electronics\nsection and turn it into images of the radio sky that are u
 sable for\nscientific analysis. It is a unique computing challenge because
  of the\ncombined demands of very high input data rates (~10 TeraByte/s)\,
  large\ncomputational requirement (around 100 PetaFLOP/s) and the need for
 \niterative algorithms requiring large working storage (around 200\nPetaBy
 te). The solution will need to wed elements of traditional\nscientific "hi
 gh-performance computing" and data-driven parallel work\ndistribution stra
 tegies.\n\nAfter reviewing the main drivers of computational requirements 
 for the\nSDP I will present the preliminary architecture of both the hardw
 are and\nsoftware that has been developed to meet those requirements. I wi
 ll then\ndescribe the ongoing work to do detailed design and development o
 f some\nof the critical parts of the system:\n\n   - The job-parallelisati
 on / data-distribution  system\n   - The model for dealing with failures a
 nd stragglers\n   - The programming model\n   - The high-performance stora
 ge subsystem
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR