Overview: Model II
The CLC Grid Integration allows jobs to be offloaded from a master server onto grid nodes using the local grid submission/queuing software to handle job scheduling and submission.A given CLC algorithm will only run on a single machine, i.e., a single job will run on one node. Each grid node employed for a given task must have enough memory and space to carry out that entire task.
The grid system uses two locations for its deployment:
- Path to CLC Grid Worker in the grid preset editor. This location is used for plugins, a grid version of the server, i.e., the code that is executed when a grid job is started, licences, libraries and more. Grid workers are redeployed in two situations: 1) When the server starts up and 2) If the configuration of one of the grid presets is changed, in which case the grid workers of all presets are redeployed. In addition, the grid workers are updated when a plugin is installed or removed.
- Shared work directory. This location is where each grid job gets its own sub directory in which it places its temporary data, e.g., the description of the job to be executed, the configurations files that the grid version of the server needs in order to setup persistence models and log files. This location is only deployed once, either when the grid job starts executing (in case of workflow jobs) or when the grid job is queued (in all other cases).