Commit a96c866f authored by Stefan Reif's avatar Stefan Reif
Browse files

Differential timing analysis

parent 7b509e64

xlap.yml

+8 −8
Original line number Diff line number Diff line
@@ -39,14 +39,14 @@ stamps: 
        Source: sender

        Thread: app_send

        Type: time

    PrrtSendPacketStart:

        Source: sender

        Thread: trans_send

        Type: none

    PrrtSendPacketEnd:

        Source: sender

        Thread: trans_send

        Type: none

#    PrrtSendPacketStart:

#        Source: sender

#        Thread: trans_send

#        Type: none

#    PrrtSendPacketEnd:

#        Source: sender

#        Thread: trans_send

#        Type: none



    PrrtTransmitStart:

        Source: sender

xlap/analyse/diff.py

0 → 100644
+39 −0
Original line number Diff line number Diff line 
import math

import numpy as np

from .util import cdf, extract_durations

from scipy import stats

import matplotlib.pyplot as plt

import xlap.analyse.latency as latency



confidence = 0.99



def remove_outliers(data):

    m = 2

    u = np.mean(data)

    s = np.std(data)

    filtered = [e for e in data if (u - 2 * s <= e and e <= u + 2 * s)]

    return filtered



def samples_are_different(sample1, sample2):

    # TODO: handle insufficient data. This case typically occurs when a duration is always zero.

    if 1 >= len(sample1) or 1 >= len(sample2) or 0 == np.std(sample1) or 0 == np.std(sample2):

        print("insufficient data")

        return 0

    a2, critical, pvalue = stats.anderson_ksamp([sample1, sample2])

    if a2 > critical[4]:

        return 1 # KS reject: samples are different

    else:

        return 0 # KS accept: samples are equal or similar



def duration_to_string(d):

    return str(d['Start']) + "->" + str(d['Stop'])



def analyse(df1, df2, config, export=False):

    for d in latency.get_durations(df1, config):

        data1 = df1[d['Stop']] - df1[d['Start']]

        data2 = df2[d['Stop']] - df2[d['Start']]

        if samples_are_different(remove_outliers(data1), remove_outliers(data2)) and samples_are_different(data1, data2):

            print(duration_to_string(d) + " has changed: "+str(np.mean(data1))+"+-"+str(np.std(data1))+" <-> "+str(np.mean(data2))+"+-"+str(np.std(data2))+" ")

        else:

            print(duration_to_string(d) + " has not changed significantly")

xlap/analyse/latency.py

+26 −0
Original line number Diff line number Diff line
@@ -178,6 +178,32 @@ def _plot_critical_regions(hdb): 
    plt.savefig("latency-criticality.pdf")

    plt.close()



def get_durations(df, config):

    hb = []



    events = [column for column in df.columns if column.endswith("_T")]



    for event1 in df[events]:

        for event2 in df[events]:

            if str(event1) == str(event2):

                continue

            if _happens_before(df, event1, event2, config):

                hb += [{'Start': str(event1), 'End': str(event2)}]



    hdb = []

    e2e = list(df['EndToEnd_D'])

    for event1 in df[events]:

        for event2 in df[events]:

            if str(event1) == str(event2):

                continue

            # if _locally_happens_directly_before(df, event1, event2, hb, config):

            if _happens_directly_before(df, event1, event2, hb):

                # compute the correlation between e2e latency and event1-event2 latency

                l3 = list(df[event2] - df[event1])



                hdb += [{'Start': str(event1), 'Stop': str(event2), 'Source': 'cfa'}]

    return hdb



def analyse(df, config):

    hb = []

xlap/command_line.py

+11 −0
Original line number Diff line number Diff line
@@ -3,10 +3,12 @@ import argparse 
from xlap.parse import evaluate, parse_config

import xlap.analyse.jitter as jitter

import xlap.analyse.latency as latency

import xlap.analyse.diff as difference



tasks = {

    "jitter": None,

    "latency": None,

    "difference": None,

    "capture": None

}
@@ -45,6 +47,15 @@ def main(): 
            df_data = evaluate(data_files["sender"], data_files["receiver"], config=config, kind=0)

            a = latency.analyse(df_data, config)

            print(a.corr.sort_values(ascending=False))

        elif command == "difference":

            df_data1 = evaluate("../prrt/out/s.csv", "../prrt/out/r.csv", config=config, kind=0)

            # sanity check:

            #df_data2 = evaluate("../prrt/out/s.csv", "../prrt/out/r.csv", config=config, kind=0)

            # same setup, different measurement run:

            #df_data2 = evaluate("../prrt/out/s+same.csv", "../prrt/out/r+same.csv", config=config, kind=0)

            # different setup:

            df_data2 = evaluate("../prrt/out/s+send.csv", "../prrt/out/r+send.csv", config=config, kind=0)

            difference.analyse(df_data1, df_data2, config)

        else:

            df_data = evaluate(data_files["sender"], data_files["receiver"], config=config, kind=0)