simulation.py

# -*- coding: utf-8 -*-
"""
    MEC_offloading.simulation
    ~~~~~~~~~~~~~~~~~~~~~~~~~

    Simulation for the MEC_offloading

    :copyright: (c) 2018 by Giorgos Mitsis.
    :license: MIT License, see LICENSE for more details.
"""

import numpy as np
import matplotlib.pyplot as plt

from parameters import *
from helper_functions import *
from game_functions import *
from server_selection_functions import *
from metrics import *
from plots import *
from create_plots import *

import time
import itertools
import dill

# Keep only three decimal places when printing numbers
np.set_printoptions(formatter={'float': lambda x: "{0:0.3f}".format(x)})

# Generate all cases
cases_setup = {
        'users': ['homo','hetero'],
        'servers': ['homo','hetero','one-dominant','two-dominant']
        }

keys, values = zip(*cases_setup.items())

# Select which case to run
cases = [{"users": "hetero", "servers": "hetero"}]
# cases = [dict(zip(keys, v)) for v in itertools.product(*values)]

for repetition in range(1000):
    print("Repetition no: " + str(repetition+1))

    results = {}
    for case in cases:

        if LOAD_SAVED_PARAMETERS == True:
            print("Loading parameters")
            infile = "saved_runs/parameters/" + case["users"] + "_" + case["servers"] + "_lr_" + "0.20"

            with open(infile, 'rb') as in_strm:
                params = dill.load(in_strm)
        else:
            # Set random parameter in order to generate the same parameters
            print("Generating new parameters")
            np.random.seed(13)
            params = set_parameters(case)

        U = params['U']
        S = params['S']
        fs = params['fs']
        c = params['c']
        b_max = params['b_max']

        start = time.time()

        # Initialize empty arrays for results
        all_server_selected = all_bytes_offloaded = all_user_utility = np.empty((0,U), int)
        all_bytes_to_server = all_prices = all_c = all_fs = all_relative_price = all_server_welfare = all_Rs = all_congestion = all_penetration = np.empty((0,S), int)
        all_probabilities = [[] for i in range(U)]

        # Get the initial values for probabilities and prices
        probabilities, prices = initialize(**params)

        for i in range(U):
            all_probabilities[i].append(probabilities[i])

        if CONSTANT_PRICING:
            # Set constant price if needed
            constant_price = np.array([1.96, 1.88, 1.94, 1.78, 1.92])
            prices = constant_price

        # Repeat until every user is sure on the selected server
        while not all_users_sure(probabilities):
            # Each user selects a server to which he will offload computation
            server_selected = server_selection(probabilities, **params)
            # Add the selected servers as a row in the matrix
            all_server_selected = np.append(all_server_selected, [server_selected], axis=0)

            # Game starts in order to converge to the optimum values of data offloading
            # Repeat until convergence for both users and servers

            if CONSTANT_OFFLOADING:
                b_old = np.ones(U) * 0.586 * b_max
            else:
                b_old = np.ones(U)

            prices_old = np.ones(S)

            converged = False
            while not converged:
                # Users play a game to converge to the Nash Equilibrium
                if CONSTANT_OFFLOADING:
                    b = b_old
                else:
                    b = play_offloading_game(server_selected, b_old, prices_old, **params)

                if CONSTANT_PRICING:
                    # Servers set their next price as they had initally set
                    prices = constant_price
                else:
                    # Servers update their prices based on the users' offloading of data
                    prices = play_pricing_game(server_selected, b, **params)

                # Check if game has converged
                converged = game_converged(b,b_old,prices,prices_old, **params)

                b_old = b
                prices_old = prices

            all_bytes_offloaded = np.append(all_bytes_offloaded, [b], axis=0)

            # Find all bytes that are offloaded to each server
            bytes_to_server = np.bincount(server_selected, b, minlength=S)
            all_bytes_to_server = np.append(all_bytes_to_server, [bytes_to_server], axis=0)

            all_prices = np.append(all_prices, [prices], axis=0)

            all_fs = np.append(all_fs, [fs], axis=0)
            all_c = np.append(all_c, [c], axis=0)

            # Calculate the welfare of the servers
            server_welfare = calculate_server_welfare(prices, bytes_to_server, **params)
            all_server_welfare = np.append(all_server_welfare, [server_welfare], axis=0)

            # Calculate the perceived utility of the users
            user_utility = calculate_user_utility(b, server_selected, prices, **params)
            all_user_utility = np.append(all_user_utility, [user_utility], axis=0)

            # Calculate the competitiveness of each server
            Rs,relative_price,congestion,penetration = calculate_competitiveness(all_bytes_to_server, all_fs, all_prices, **params)

            all_Rs = np.append(all_Rs, [Rs], axis=0)
            all_congestion = np.append(all_congestion, [congestion], axis=0)
            all_penetration = np.append(all_penetration, [penetration], axis=0)
            all_relative_price = np.append(all_relative_price, [relative_price], axis=0)

            # Update the probabilities
            probabilities = update_probabilities(Rs, probabilities, server_selected, b, **params)

            for i in range(U):
                all_probabilities[i].append(probabilities[i])

        for i in range(len(all_probabilities)):
            all_probabilities[i] = np.array(all_probabilities[i])
        all_probabilities = np.array(all_probabilities)

        end = time.time()
        running_time = end - start
        print("Time of simulation:")
        print(running_time)

        # Keep results in a dictionary in order to save and plot them
        key = case["users"] + "_" + case["servers"]
        results[key] = {
            "all_bytes_offloaded": all_bytes_offloaded,
            "all_server_selected": all_server_selected,
            "all_prices": all_prices,
            "all_bytes_to_server": all_bytes_to_server,
            "all_server_welfare": all_server_welfare,
            "all_user_utility": all_user_utility,
            "all_Rs": all_Rs,
            "all_relative_price": all_relative_price,
            "all_congestion": all_congestion,
            "all_penetration": all_penetration,
            "all_fs": all_fs,
            "all_c": all_c,
            "all_probabilities": all_probabilities,
            "running_time": running_time
            }

        # Save parameters and results
        if SAVE_PARAMETERS == True:
            if CONSTANT_PRICING == True:
                outfile = "saved_runs/parameters/" + case["users"] + "_" + case["servers"] + "_lr_" + "{0:.2f}".format(params["learning_rate"]) + "_constant-pricing"
            else:
                outfile = "saved_runs/parameters/" + case["users"] + "_" + case["servers"] + "_lr_" + "{0:.2f}".format(params["learning_rate"])

            with open(outfile, 'wb') as fp:
                dill.dump(params, fp)

        if SAVE_RESULTS == True:
            if CONSTANT_PRICING == True:
                outfile = 'saved_runs/results/individual/' + case["users"] + "_" + case["servers"] + "_lr_" + "{0:.2f}".format(params["learning_rate"]) + "_constant-pricing" + "_rep_" + str(repetition+1)
            else:
                outfile = 'saved_runs/results/individual/' + case["users"] + "_" + case["servers"] + "_lr_" + "{0:.2f}".format(params["learning_rate"]) + "_rep_" + str(repetition+1)

            with open(outfile , 'wb') as fp:
                dill.dump(results[key], fp)

# Create the plots
# create_plots(results, cases, params)