Workers and Groups

Workers

To specify the amount of workers (the default is 4) to run a job:

execution.workerCount = 16;

4 is the default. There is a maximum of 8192. If you specify a worker count higher than the number of available eligible workers, your job will block remain queued until the specified amount of workers becomes available.

Each worker has a score based on their previous history of job participation. Workers with higher scores are more likely to be chosen to participate in a job.

WebGPU

You can require that a worker has WebGPU support by setting the gpu flag:

execution.gpu = true;

When this flag is set, only workers with WebGPU support will be chosen to participate in the job.

Private Workers

When this flag is set, only your own workers will be chosen to participate in the job. You can bring up your own workers by connection multiple tabs or browsers to PeerMR. This mode ensures data privacy as no other workers will execute any parts of the job.

In addition., while developing a job, you may want to test it with a small number of your own workers before running it with a larger group of workers.

Private mode is enabled by setting the privateMode flag:

execution.privateMode = true;

Groups

You may wish to execute your job multiple times at once and compare the results. This is easily done with groups. A group count of 2 for example will execute the job twice, separating the intermediate and final output into groups A and B.

To specify the amount of groups for your job:

execution.groupCount = 2;