Executing Jobs on Your Compute Service
A Job is a calculation which TRAC orchestrates on your Compute Service. The platform integrates with a range of compute technologies. For cloud deployments we provide a Kubernetes pattern as standard supporting both single node and distributed (Spark) capabilities.
JOB INSTRUCTION
A Job record consists of a packaged instruction describing the calculation plus the data asset(s) which it generated. Between those two elements is the calculation that TRAC orchestrates on your Compute Service using the Job instruction.
THE AUDIT RECORD
For the calculation to run, each Job instruction must provide a complete description of the calculation (the Flow) and all its inputs (Models and Data). The Job instruction therefore doubles as the Job’s permanent audit record.
TWEAK AND REPEAT
Any Job can be re-run and TRAC will return the same result, guaranteed. You can also find, tweak and re-peat a previously executed Job, to isolate the impact of model, portfolio or parameter changes.
