log in  |  register  |  feedback?  |  help  |  web accessibility
Logo
Programming and Systems Abstractions for Democratizing Data-Intensive Computing
Tuesday, April 30, 2013, 2:00-3:00 pm Calendar
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract
Data-intensive computing promises to greatly enrich our understanding of natural, social and mechanized phenomena, yet our progress is often stunted by an implementation bottleneck rather than algorithmic creativity. Scientists, analysts and researchers suffer a heavy engineering burden in transitioning on-paper, or in-the-lab efforts into production-grade software artifacts for use in data-driven experimentation. 

This talk spans two ongoing efforts at Johns Hopkins to design programming and systems abstractions that enable computational scientists to build powerful and scalable tools in small teams, rather than with many tens or hundreds of engineers. The first, a bottom-up effort, is K3, a general event-driven programming language that decouples of algorithms concerns from implementation and execution concerns. K3 users can declaratively specify data structures, as well as parallelization and resiliency features of their programs amongst other properties.

Next, a top-down effort, is our co-operative analysis framework that enables the easy hybridization of analysis algorithms from best-of-breed approaches. Our framework is agnostic to the form of co-operation (i.e., across different data representations or resolutions), provides efficient data movement and online operation, and draws inspiration from applications in molecular dynamics and computational drug design.
This talk is organized by Abdul Quamar