|
Abstract:
|
Compute-intensive statistical tasks often can be factored into repeating many small, independent tasks. Examples include repetitive analyses on partitions of large databases, bootstrapping, simulation, and Bayesian sampling. Such tasks are potentially amenable to grid computing, where the small tasks are farmed to multiple "server nodes." Issues to be discussed: What sorts of realistic compute-intensive problems do people run across? Can grid computing be expected to help? What hardware and software are required? How do you go about turning a one-machine task into a multiple-machine task? The leaders have experience with statistical grid computing, mainly using SAS software. They will be prepared to introduce the methodology and talk about their successes (and otherwise!). Participants should come with real statistical problems in mind that may benefit from grids.
|
