Skip to contents

Stochastic Network Models on High-Performance Computing Systems

Source: R/netsim_hpc.R
netsim_hpc.Rd

Simulates stochastic network epidemic models for infectious disease dynamics in parallel.

Usage

netsim_hpc(
  x,
  param,
  init,
  control,
  cp.save.int = NULL,
  save.min = TRUE,
  save.max = FALSE,
  compress = TRUE,
  verbose = TRUE
)

Arguments

x

Character vector containing the file path of an Rdata file where an object of class netest is stored. Alternatively, if restarting a previous simulation, this may be a file path for an object of class netsim.

param

Model parameters, as an object of class param.net.

init

Initial conditions, as an object of class init.net.

control

Control settings, as an object of class control.net.

cp.save.int

Check-pointing save interval, which is used to specify how often intermediate data should be saved out to disk. When a job has been check-pointed, it will resume automatically at the last saved time step stored on disk. If set to NULL, then no intermediate data storage will occur.

save.min

Argument passed to savesim.

save.max

Argument passed to savesim.

compress

Matches the compress argument for the save function.

verbose

If FALSE, suppress all output messages except errors.

Details

This function provides a systematic method to running stochastic network models in parallel on high-performance computing systems.

The main purpose of using netsim_hpc is for a standardized checkpointing method. Checkpointing is defined as incrementally saving simulation data for the purpose of reloading it if a simulation job is canceled and restarted. If checkpointing is not needed, users are advised to run their models directly with the EpiModel::netsim function.

This function performs the following tasks:

  1. Check for the existence of checkpointed data, using the check_cp function. If CP data are available, a checkpointed model will be run, else a new model will be run.

  2. Create a checkpoint directory if one does not exist at "data/simsimno". This and the related checkpointing functions will not occur if cp.save.int is set to NULL.

  3. Sets the checkpoint save interval at the number of time steps specified in cp.save.int.

  4. Resets the initialize module function to initialize_cp if in checkpoint state.

  5. Run the simulation, either new or checkpointed, with a call to EpiModel::netsim.

  6. Save the completed simulation data, using the functionality of savesim.

  7. Remove the checkpointed data and file directory created in step 1, if it exists.

The x argument must specify a file name in a character string, rather than a netest or netsim class object directly. This is mainly for efficiency purposes in running the models in parallel.

If save.min and save.max are both set to FALSE, then the function will return rather than save the output EpiModel object.