About & News
   Earth Science
   Computational Science
      Heterogeneous Computing
      Robust Middleware
      Scalable Algorithms
      Desktop HPC
      Parallel Matlab
      Distributed ensembles
      Web and Grid Services
      Sensor Integration
   Technical Blueprints
   Example Applications
Getting Started
Sites
Available software
Status
People
FAQ
Mailing Lists
Quick Links
Contact Us
Sponsors
Search

MIT logo

Scalable algorithms and applications

Many of the most compelling computaitonal earth science problems require efficient use of computational resources. The ACES system will be used for further developing scalable numerical techniques for planetary scale systems. Some examples of planned ACES work in this area includes:

Ongoing development of scalable medium level complexity coupled modeling systems.

Scalable approaches to gridding the sphere.

High-resolution tracer transport modeling with scalable I/O: Many fluid Earth problems can be studied by using precomputed flow fields to transport one or more tracer. Numerically these methods can be very efficient because certain wave phenomena that restrict numerical simulation time steps are absent from transport models. However, high resolution transport fields require storing flow fields at high temporal frequency resulting in potentially multi-terabyte datasets. To retain efficiency transport models must, therfore, stream I/O in parallel in a scalable manner. The network fabric in ACES will be used to further research techniques that are already being applied at MIT to these problems.