Presentation: "Fine Grained Coordinated Parallelism in a Real World Application"

Time: Tuesday 16:50 - 17:50

Location: Salon A-B

Abstract:
For certain types of problems, there is no better solution than a highly tuned, shared-memory parallel algorithm. This talk will showcase code that achieves  concurrency via lock free algorithms. We'll explore how complex aggregation can be performed using fine grained coordinated parallelism. To accomplish this goal, we'll discuss advanced data structures, like a brand new lock-free hashmap with interesting features such as concurrent resizing. By using some of Java's less commonly used API, such as AtomicFieldUpdaters, we'll demonstrate how a real world application can be written to not only work with a dozen cores, but be ready for a 1000.  

There is a lot of angst about writing concurrent code. The message of this talk is one of courage and hope without leaning on some future language or technology. Writing concurrent code requires a different mindset. The approaches detailed in this talk will hopefully serve as an introduction that will strengthen the audience's resolve to explore multi-threaded programming.

Mohammad Rezaei, PhD, Vice President, Technical Architect, Goldman Sachs

 Mohammad  Rezaei
Mohammad Rezaei is the co-CTO for the Controllers Technology business unit at Goldman Sachs. He's had extensive experience writing high performance Java code in a variety of environments, from partitioned/concurrent transactional systems to large memory systems that require lock free algorithms for maximum throughput. Mohammad hold a B.S. in Computer Science from University of Pennsylvania and a Physics PhD from Cornell University.