Deep Learning for Numerical Applications with SAS

Chapter 4: Many-Task Computing

A Taxonomy for Parallel Programs

Tasks Are the New Threads

What Is a Task?

Inputs and Outputs

Immutable Inputs

What Is a Job Flow?

Examples of Job Flows

Mutable Inputs

Task Revisited

Partitioning

Federated Areas

Persistent Area

Caveats and Pitfalls

Not Declaring Your Inputs

Not Treating Your Immutable Inputs as Immutable

Not Declaring Your Outputs

Performance of Grid Scheduling

Data-Object Pooling

Portable Learning

Conclusion

 

In this chapter, we take a slight detour from deep learning (DL) and venture into some supercomputing technologies, namely many-task computing (MTC). We take this detour because we want to take advantage of the latest developments in that field and apply them to the development and deployment ...