10.2. Using the virtual machine abstraction layer

This tutorial explains the basics of the virtual machine abstraction layer. Currently only MPI is supported, but one could add an abstraction of virtually any parallel processing platform (i.e. in the past PyOPUS was using PVM in much the same manner as MPI today). The tutorial explains how to retrieve the virtual machine configuration, how to spawn functions of remote processors, and how to establish communication between tasks. The tutorial also explains how to perform file mirroring so that all tasks run in an identical environment that is established on a local storage to maximize performance. Finally an example is given that outsources some tasks and collects their results.

• 10.2.1. Using MPI with PyOPUS
• 10.2.2. Retrieving the machine/cluster configuration
• 10.2.3. Spawning a task on a (remote) host
• 10.2.4. Measuring the message delay and throughput
• 10.2.5. Data mirroring
• 10.2.6. Removing old local storage folders
• 10.2.7. Spawning remote tasks and collecting results