batch-shipyard/recipes/NAMD-TCP
Fred Park 9736cea593
Update NAMD-TCP recipe
- Fix connection issues
2019-07-09 21:46:34 +00:00
..
config Update NAMD-TCP recipe 2019-07-09 21:46:34 +00:00
docker Allow NAMD-TCP recipe to be run on one node 2017-02-14 09:24:17 -08:00
README.md Update NAMD-TCP recipe 2019-07-09 21:46:34 +00:00

README.md

NAMD-TCP

This recipe shows how to run NAMD on Linux using the Charm++ runtime (as opposed to pure MPI) over TCP/IP-connected machines in an Azure Batch compute pool. Regardless of the underlying parallel/distributed programming paradigm, execution of this distributed workload requires the use of multi-instance tasks when run across multiple nodes.

Configuration

Please see refer to this set of sample configuration files for this recipe.

Pool Configuration

The pool configuration should enable the following properties:

  • inter_node_communication_enabled must be set to true
  • max_tasks_per_node must be set to 1 or omitted

Global Configuration

The global configuration should set the following properties:

  • docker_images array must have a reference to the NAMD-TCP image. This can be alfpark/namd:2.11-tcp which is published on Docker Hub.

Jobs Configuration

The jobs configuration should set the following properties within the tasks array which should have a task definition containing:

  • docker_image should be the name of the Docker image for this container invocation, e.g., alfpark/namd:2.11-tcp
  • additional_docker_run_options should have a sequence item of --net=host to specify host networking. This is needed to open ports for communication via TCP between containers.
  • command should contain the mpirun command. If using the sample NAMD-TCP image provided, "/sw/run_namd.sh <benchmark> <steps> <ppn>" can be used to run the included benchmarks:
    • <benchmark> is the benchmark to run: apoa1 or stmv
    • <steps> is the number of steps to execute
    • <ppn> is the number of cores on each compute node. This is optional and, if omitted, will be determined dynamically.
  • multi_instance property must be defined for NAMD tasks spanning multiple nodes.
    • num_instances should be set to pool_specification_vm_count_dedicated, pool_vm_count_low_priority, pool_current_dedicated, or pool_current_low_priority
    • coordination_command should be unset or null. For pools with native container support, this command should be supplied if a non-standard sshd is required.
    • resource_files array can be unset or empty

To run this example on just one node, you can omit the multi_instance property altogether.

Dockerfile and supplementary files

The Dockerfile for the Docker image can be found here. Please note that you must agree with the NAMD license before using this Docker image.