Conventional wisdom suggests that making large-scale distributed computations fast requires minimizing the latency of individual operations in the computation. In this talk I will discuss a system called Grappa that takes the opposite view. Grappa tolerates latency by exploiting application parallelism to achieve overall higher throughput.
Grappa is a modern take on software distributed shared memory for in-memory data-intensive applications. Grappa enables users to program a cluster as if it were a single, large, non-uniform memory access machine. Performance scales up even for applications that have poor locality and input-dependent load distribution, as long as sufficient parallelism is available.
Jacob Nelson is a Postdoctoral Research Associate in the Department of Computer Science and Engineering at the University of Washington. Jacob’s research explores new software and hardware techniques to accelerate applications in big data and high-performance computing. Jacob defended his Ph.D. at the University of Washington in 2014 working with Luis Ceze, Mark Oskin, and Simon Kahan.