Services are unique workflows where there is an iterative component on the server. A typical service workflow looks like the following:
A client submits a new service request to the server.
A service is created on the server and placed in the service queue.
A service iteration is called that will spawn new tasks.
The service waits until all generated tasks are complete.
The service repeats 3 and 4 until the service iterations are complete.
The service cleans intermediate data, finalizes the data representation, and marks itself complete.
The TorsionDrive service will be used as an example to illuminate the above steps. The TorsionDrive service optimizes the geometry of a biomolecule at a number of frozen dihedral angles to provide an energy profile of the rotation of this dihedral bond.
Consider the service using a concrete example of scanning the hydrogen peroxide dihedral:
A client submits a task to scan the HOOH molecule dihedral every 90 degrees as a service.
The service is received by the server, and the first 0-degree dihedral geometry optimization Task is spawned.
The service waits until the 0-degree Task is complete, and then generates 90 and -90-degree tasks based off this 0-degree geometry.
The service waits for the two new tasks to complete and spawns 0 and 180-degree tasks based on the 90 and - 90-degree geometries.
The service waits for the 90- and -90-degree tasks to complete. Then it builds its final data structure for user querying and marks itself complete.
The service technology allows the
FractalServer to complete very complex
workflows of arbitrary design. To see a pictorial representation of this
process, please see the
flowchart showing the pseudo-calls when a
service is added to the