BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Scaling transaction processing through data-oriented execution - I
 ppokratis Pandis\, Carnegie Mellon University
DTSTART:20110704T090000Z
DTEND:20110704T100000Z
UID:TALK31925@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Transaction processing is one of the most important and challe
 nging data management workloads. Unfortunately\, while hardware technology
  has undergone major advancements over the past decade\, the core architec
 ture of transaction processing systems has remained largely unchanged. The
  number of cores on a chip grows exponentially\, following Moore's Law\, a
 llowing for an ever increasing number of transactions to execute in parall
 el. As the number of concurrently-executing transactions increases\, conte
 nded critical sections become scalability burdens.\nIn this talk we first 
 categorize the different types of critical sections a system executes duri
 ng transaction processing and argue that not all of them are threats to sc
 alability. Then\, we identify the thread-to-transaction work assignment po
 licy as the primary cause of contention of conventional shared-everything 
 systems\, and argue for the adoption of a data-oriented (or logically-part
 itioned) transaction execution design. \nFinally\, we present the design o
 f a transaction processing system under physical/logical partitioning (PLP
 ) which achieves a dual goal. Like a physically-partitioned system the maj
 ority of data accesses occur in a single-threaded environment which render
 s the majority of unscalable critical sections unnecessary\; like a logica
 lly-partitioned system it maintains the ACID properties\, locking is distr
 ibuted without resorting to distributed transactions\, and load balancing 
 is inexpensive because almost no data movement is necessary. Evaluation of
  a prototype implementation of PLP on multicore hardware demonstrates that
  PLP attains up to 6x higher throughput than a conventional system when ru
 nning synthetic and real-world OLTP workloads.
LOCATION:Small lecture theatre\, Microsoft Research Ltd\, 7 J J Thomson Av
 enue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR