SeD requirements for workflow scheduling

The multi-workflow schedulers (Multi-HEFT, Multi-AgingHEFT and FOFT) use information provided by the SeDs to be able to run the heuristic. So the SeD programmer must provide the required data in the estimation vector that is sent back to the client for each submit request.

The following fields in the estimation vector must be filled in:

1. The TCOMP field must contain the estimation of the computation time for the job (in milliseconds). This can be done using the diet_estimate_comptime(estVector_t ev, double value) method within the performance evaluation function.
2. The USERDEFINED field must contain the estimation of the earliest finish time (in milliseconds from the time of the current submit request) for the job. To compute this value, the SeD programmer can use the API method diet_estimate_list_jobs(...) to retrieve the estimated values of computation time for all the jobs active on the SeD (ie jobs waiting for solve or jobs being solved). There is an example of this EFT calculation in the simple case of sequential SeD available in the workflow examples (scalar_server or file_server).