Template class for algorithm

[rantahar]

• Move shared functions
• Use the template in each current algorithm
• Use network and disk bound parameters
• Allow user defined task parameters
• Store task run data in controller and communicate to scheduler
  • Store start time, duration, input file size and number
• Break down the assign_task function
  • can_assign_on_node function to check if a task can be assigned to a node
  • Function that subtracts task resources from node
  • Save inputs and outputs to a file for ML training
• [ ]

[nguu0123]

I think this can be finished fast so if you have time after high-priority issue, you can implement this

[rantahar]

Disk storage format for the input and output format for the can_assign_on_node function

• Inputs:
  • Dictionary

  {
           'resources: {
                      'memory': int,
                      'cpus': []  # Example: ["core_1", "core_2", ...],
                      'private_disk': int,
                      'shared_disk': int,
                      'free': bool,
           },
          'task_requirements': {
                    'execution_time': int,
                    'ranks': int,
                    'cpus': int,
                    'network_bound': bool,
                    'disk_bound': bool,
          }
  }
  

• Outputs:
  • Binary classification