4. Usage Examples

Usage examples are provided in the notebook mosaic_schedulers/common/examples/scenarios/example_scenarios.ipynb, reproduced below.

from mosaic_schedulers.schedulers.tv_milp import MOSAICSolver
from mosaic_schedulers.schedulers.ti_milp import MOSAICTISolver
from mosaic_schedulers.schedulers.ti_milp_heft import Scheduler as MOSAICTIHEFTSolver

from mosaic_schedulers.common.plotting import MOSAICplotter
import matplotlib.pyplot as plt
import json

import static_demo

%matplotlib notebook

4.1. Create a problem instance

Create a sample input to the problem by choosing a demo type among: - “selfish”: all agents work in isolation - “node_assist”: nodes can help each other - “base_station_assist”: nodes have access to a powerful base station for computation but can’t assist other nodes - “assembly_line”: underemployed nodes act both as relays and as computational assists

demo_input = static_demo.demo_input(demo_type="assembly_line", network_size='large',time_horizon=30)

JSONinput, TaskColors = static_demo.create_json_description(demo_input)

# Cutoff on the MILP solver time
time_limit = 120

For additional control on the problem inputs, you may want to call the function create_problem_input with parameters - science_nodes_number: the number of agents in a science zone - assist_nodes_number: the number of agents not in a science zone - inter_puffer_bw: the bandwidth between agents - base_station_bw: the bandwidth to-from a base station - assembly_chain_bw_hop1: the bandwidth between selected nodes in a science zone and their counterparts not in a science zone - assembly_chain_bw_hop2: the bandwidth between selected nodes not in a science zone and the base station - time_horizon: the time horizon of the optimization - agent: what agent is calling the solver (useful if the problem is partitioned)

For instance,

static_demo.demo_input(demo_type="assembly_line", network_size='large',time_horizon=30)

internally calls

create_problem_input(science_nodes_number=4, assist_nodes_number=4, inter_puffer_bw=0.01, base_station_bw=0.01, assembly_chain_bw_hop1=5.5, assembly_chain_bw_hop2=0.5, time_horizon=30, agent='puffer1',):

4.2. Solve with a time-varying MILP solver

# Test with a TV solver
Scheduler = MOSAICSolver.JSONSolver(JSONinput, TimeLimit=time_limit)
Schedule = Scheduler.schedule()

# print("TV SCHEDULE:\n", Schedule)
if Schedule is not None:
    MOSAICplotter.SchedulePlotter(Schedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(Schedule,TaskColors=TaskColors)
    evaluation = static_demo.evaluate_output(JSONinput, Schedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 4/4
 - Stored: 1/4

4.3. Allocate tasks with a “time-invariant” solver

# # Test with a TI solver
TIScheduler = MOSAICTISolver.JSONSolver(JSONinput,solver='CPLEX',TimeLimit=60)
TISchedule = TIScheduler.schedule()
# print("TI SCHEDULE:\n", TISchedule)
if TISchedule is not None:
    MOSAICplotter.TaskAllocationPlotter(TISchedule,TaskColors=TaskColors)
    plt.show()

# Test with TI-HEFT

TIHScheduler = MOSAICTIHEFTSolver(JSONinput,solver='CPLEX',TimeLimit=time_limit)
TIHSchedule = TIHScheduler.schedule()
#print("TI HEFT SCHEDULE:\n", TIHSchedule)

if TIHSchedule is not None:
    MOSAICplotter.SchedulePlotter(TIHSchedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(TIHSchedule,TaskColors=TaskColors)
    plt.show()
    _ = static_demo.evaluate_output(JSONinput, TIHSchedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 4/4
 - Stored: 4/4

4.4. Selfish agents

demo_input = static_demo.demo_input(demo_type="selfish", network_size='large',time_horizon=30)
JSONinput, TaskColors = static_demo.create_json_description(demo_input)
# Test with a TV solver
Scheduler = MOSAICSolver.JSONSolver(JSONinput, TimeLimit=time_limit)
Schedule = Scheduler.schedule()

# print("TV SCHEDULE:\n", Schedule)
if Schedule is not None:
    MOSAICplotter.SchedulePlotter(Schedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(Schedule,TaskColors=TaskColors)
    _ = static_demo.evaluate_output(JSONinput, Schedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 0/4
 - Stored: 0/4

4.5. Node Assist

demo_input = static_demo.demo_input(demo_type="node_assist", network_size='large',time_horizon=30)
JSONinput, TaskColors = static_demo.create_json_description(demo_input)
# Test with a TV solver
Scheduler = MOSAICSolver.JSONSolver(JSONinput, TimeLimit=time_limit)
Schedule = Scheduler.schedule()

# print("TV SCHEDULE:\n", Schedule)
if Schedule is not None:
    MOSAICplotter.SchedulePlotter(Schedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(Schedule,TaskColors=TaskColors)
    _ = static_demo.evaluate_output(JSONinput, Schedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 4/4
 - Stored: 0/4

4.6. Base Station Assist

demo_input = static_demo.demo_input(demo_type="base_station_assist", network_size='large',time_horizon=30)
JSONinput, TaskColors = static_demo.create_json_description(demo_input)
# Test with a TV solver
Scheduler = MOSAICSolver.JSONSolver(JSONinput, TimeLimit=time_limit)
Schedule = Scheduler.schedule()

# print("TV SCHEDULE:\n", Schedule)
if Schedule is not None:
    MOSAICplotter.SchedulePlotter(Schedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(Schedule,TaskColors=TaskColors)
    _ = static_demo.evaluate_output(JSONinput, Schedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 3/4
 - Stored: 2/4

4.7. Assembly Line

demo_input = static_demo.demo_input(demo_type="assembly_line", network_size='large',time_horizon=40)
JSONinput, TaskColors = static_demo.create_json_description(demo_input)
# Test with a TV solver
Scheduler = MOSAICSolver.JSONSolver(JSONinput, TimeLimit=time_limit)
Schedule = Scheduler.schedule()

# print("TV SCHEDULE:\n", Schedule)
if Schedule is not None:
    MOSAICplotter.SchedulePlotter(Schedule,TaskColors)
    MOSAICplotter.TaskAllocationPlotter(Schedule,TaskColors=TaskColors)
    _ = static_demo.evaluate_output(JSONinput, Schedule,verbose=True)
else:
    print("Problem infeasible!")

Samples:
 - Taken: 4/4
 - Analyzed: 4/4
 - Stored: 4/4