BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Turbocharging Rack-Scale In-Memory Computing with Scale-Out NUMA -
  Boris Grot - University of Edinburgh
DTSTART:20160504T100000Z
DTEND:20160504T110000Z
UID:TALK66084@talks.cam.ac.uk
CONTACT:44515
DESCRIPTION:Web-scale online services mandate fast access to massive quant
 ities of data. In practice\, this is accomplished by sharding the datasets
  across a pool of servers within a datacenter and keeping each shard in th
 e servers' main memory to avoid long-latency disk I/O. Accesses to non-loc
 al shards take place over the datacenter network\, incurring communication
  delays that are 20-1000x greater than accesses to local memory. In this t
 alk\, I will introduce Scale-Out NUMA -- a rack-scale architecture with an
  RDMA-inspired programming model that eliminates chief latency overheads o
 f existing networking technologies and reduces the remote memory access la
 tency to a small factor of local DRAM. I will overview key features of Sca
 le-Out NUMA and will describe how it can bridge the semantic gap between s
 oftware and hardware through integrated support for atomic object reads. 
LOCATION:Small Lecture Theatre\, Microsoft Research Ltd\, 21 Station Road\
 , Cambridge\, CB1 2FB
END:VEVENT
END:VCALENDAR