Jacob Nelson Grappa: faster data-intensive applications through latency tolerance

Date of Presentation: 
Wednesday, May 7, 2014
Faculty Sponsor: 
Martine de Cock
2014 Spring
Research Focus: 

Jacob Nelson - Grappa - data framework for graph applications

In this talk, I will present Grappa, a new open-source platform for accelerating in-memory data-intensive applications on commodity clusters.


Grappa enables users to program a cluster as if it were a single, large, non-uniform memory access (NUMA) machine. Performance scales up with increasing concurrency even for applications that have poor locality and input-dependent load distribution. Grappa obtains performance by exploiting abundant parallelism in applications, often delaying individual tasks to improve overall throughput. For example, Grappa moves execution to the data for improved memory bandwidth; multiplexes thousands of tasks per core to tolerate long latency operations and overlap communication and computation; and aggregates small network messages into larger packets to better utilize network bandwidth.


I will discuss the design and use of Grappa, as well as our recent experience implementing a subset of the GraphLab API as well as an execution engine for the Raco relational query compiler.




Jacob Nelson is a final-year PhD student in the Department of Computer Science and Engineering at the University of Washington in Seattle. His current research focus is on hardware and software systems to accelerate data-intensive applications.