![[logo] US EPA](http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
How SuperMUSE Works
 SuperMUSE Master Console  Layout of Network and KVM switches used in SuperMUSE to assign work and manage individual PCs |
- An individual Model Tasker (Model Dependent) is built for each modeling system.
- When launched, the Model Tasker registers with a CPU allocator (Model Independent).
- The CPU allocator oversees multiple active model taskers
- In Windows environments, the Tasker Client (Model Independent) executes DOS commands in batch files delivered by the Model Tasker.
- The Tasker Client announces availability (one of many PCs)
- The Tasker Client is assigned to a Model Tasker
- The Tasker Client requests a job from its assigned Model Tasker.
- The Tasker Client will perform Job X (a single task line in a Model Tasker)
- The Tasker Client will send warnings, errors, and results to a Data Server project file area that is identified by the Model Tasker
- The Tasker Client then indicates that Job X is completed
- The Tasker Client will then announce it's availability to the CPU Allocator, where it is randomly assigned again to some active Model Tasker.
- All inter-PC communication is handled through use of the TCP/IP network protocol.
This cycle occurs for each PC, so many model simulations are running in parallel. With an average model runtime of 2 minutes, SuperMUSE can run over 4 million simulations/month.
SuperMUSE Parallel Management Software Toolset (PDF, 17" x 11", 1 pp., 106 KB, about PDF), a diagram of this cycle for Windows environments.
Conceptual Layout of SuperMUSE Hardware and Software, a diagram of the conceptual layout of SuperMUSE hardware and software, along with its relationship to model software designed for stand-alone PCs