TomographySystem.py

import numpy as np
import pandas as pd
from scipy.stats import pearsonr

from MeasurementsHelper import *


ROUTING_MATRIX = np.array([[1, 1, 0], [1, 0, 1], [1, 1, 1]])


def get_perf_object(metric):
    if metric is PerfMetric.THROUGHPUT: return ThroughputPerf
    if metric is PerfMetric.LOSS_RATIO: return LossPerf
    if metric is PerfMetric.RETRANSMISSION_RATIO: return RetransmissionPerf
    if metric is PerfMetric.DELAY: return DelayPerf

    
def get_good_state(perf, threshold, metric):
    return get_perf_object(metric).is_state_good(perf, threshold)
    
    
def solve_system(y_vals):
    return 10 ** np.linalg.solve(ROUTING_MATRIX, np.log10(y_vals))


# V.I.P Note: a tomography system consists of only two paths and not more
class TomographySystem:
    
    def __init__(self, path_pair_perf, metric, interval_size, threshold):
        self.path_pair_perf = path_pair_perf
        self.interval_size = interval_size
        self.threshold = threshold
        self.metric = metric
        
        # compute performances at valid intervals
        self.sys_perf = path_pair_perf.compute_perfs(interval_size)
        self.sys_perf = self.sys_perf[(self.sys_perf.perf_p1 != 0) | (self.sys_perf.perf_p2 != 0)]
        
        # get the path-sets good states
        self.sys_perf['state_p1'] = self.sys_perf['perf_p1'].apply(lambda val: get_good_state(val, self.threshold, self.metric))*1
        self.sys_perf['state_p2'] = self.sys_perf['perf_p2'].apply(lambda val: get_good_state(val, self.threshold, self.metric))*1
        self.sys_perf['state_p1p2'] = self.sys_perf['state_p1'] * self.sys_perf['state_p2']
        
        # compute the correlation coefficient
        self.corr_coef = pearsonr(self.sys_perf['state_p1'], self.sys_perf['state_p2'])[0]
        
        # compute the paths probabilities 
        y_vals = [self.sys_perf['state_p1'].mean(), self.sys_perf['state_p2'].mean(), self.sys_perf['state_p1p2'].mean()]
        self.pathsets_probs = {
            'prob_p1': to_pct(y_vals[0]), 'prob_p2': to_pct(y_vals[1]), 'prob_p1p2': to_pct(y_vals[2])
        }
        
        # solve for link probabilities
        x_vals = solve_system(y_vals)
        self.links_inferred_probs = {
            'prob_lc': to_pct(x_vals[0]), 'prob_lnc1': to_pct(x_vals[1]), 'prob_lnc2': to_pct(x_vals[2])
        }