BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Breakthroughs in security\, efficiency\, and performance with on-d
 ie hetero-processing: How Hardware-Software co-design is redefining data m
 anagement
DTSTART:20160429T120000
DTEND:20160429T130000
DTSTAMP:20260502T134507Z
UID:fb2b33f356db5b385f49b90156f6ae2647a2ea25483aef8aa3c46883
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Garret Swart\, Oracle\nThe Oracle database is the world's l
 eading data management product.  Soon after the acquisition of Sun Micros
 ystems in 2010\, engineers from Oracle and Sun started design of a new cat
 egory of microprocessor that processes data several times faster\, many ti
 mes more efficiently\, and qualitatively safer.  This goal cannot be achi
 eved by running software unchanged -- we needed to design new hardware and
  write new software to utilize it.  The first step in this transformation
  was released as the SPARC M7 processor last year.  The approach we took\
 , and are taking\, exploits the following ideas:\n• Big is better:  Lar
 ge scale computing systems give access to huge amounts of data without the
  costs of moving data between systems.  This allows for larger tables\, b
 igger sorts\, fatter graphs\, and more cloud tenants sharing the the same 
 resource pool on SPARC systems that scale linearly in cost and performance
  from 8 to 512 cores\, 64 to 4096 threads.\n• Secure is better:  Cache-
 line level memory access checking allows our instrumented memory allocator
 s to manage memory at production speed while detecting bugs and reporting 
 attacks in real time.\n• Information Density is better:  With hardware 
 designed for scanning n-gram compressed\, bit packed\, dictionary and run-
 length encoded columnar data at full memory bandwidth\, we make maximal us
 e of every bit stored and every cache line transferred over the memory cha
 nnels with no impact on performance.\n• Fast is better:  With hardware 
 support for running database operators on specialized streaming processors
 \, we drive the memory channels at maximum rate\, freeing up power and cor
 es for running user computations on the result of these operators.\n• Co
 nnected is better:  Integrating EDR InfiniBand on-chip and on-board with 
 low-latency\, high-throughput\, one-sided networking.\n• Portable is bet
 ter:  By supporting platform independent acceleration APIs inside the dat
 abase we support a wide variety of acceleration techniques and give applic
 ations and query planners the information to make the best use of the avai
 lable hardware.\n• Integrated is better:  By supporting and acceleratin
 g multiple storage types (In-memory\, NVMe\, Exadata\, NFS\, HDFS\, Fibre 
 Channel)\, data formats (row major\, column major\, graph\, JSON\, spatial
 \, MIME\, Hive)\, algorithms\, query languages\, network protocols\, and h
 ardware platforms in a single product\, we optimize\, share resources\, in
 crease usability and reduce the cost and the cognitive load in acquiring\,
  storing\, securing and understanding data.\nIn this talk\, I will describ
 e the experience that drives our acceleration priorities\, the constraints
  and joys of the hardware-software co-design process\, the HW features tha
 t resulted\, and how software engineers have exploiting these features in 
 ways we expected and ways we didn't.  The industry has taken this approac
 h used in the acceleration of linear algebra and computer graphics\, and i
 n this talk we'll see how we apply it to data processing but with changes 
 to match the lower compute density of the problem space.
LOCATION:BC 410 https://plan.epfl.ch/?room==BC%20410
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR