RTN'18 dump

xlap/analyse/diff.py

@@ -29,7 +29,7 @@ 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):
+    for d in latency.get_durations_2([df1, df2], 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):

xlap/analyse/e2e.py

+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
+
+# Taken from: http://composition.al/blog/2015/11/29/a-better-way-to-add-labels-to-bar-charts-with-matplotlib/
+def autolabel(rects, ax, labels):
+    # Get y-axis height to calculate label position from.
+    (x_left, x_right) = ax.get_xlim()
+    x_width = x_right - x_left
+
+    for i, rect in enumerate(rects):
+        width = rect.get_width()
+        color = "black"
+        align = "left"
+
+        # Fraction of axis height taken up by this rectangle
+        p_width = (width / x_width)
+
+        # If we can fit the label above the column, do that;
+        # otherwise, put it inside the column.
+        if p_width > 0.50:  # arbitrary; 95% looked good to me.
+            label_position = width - (x_width * 0.01)
+            color = "white"
+            align = "right"
+        else:
+            label_position = width + (x_width * 0.01)
+
+        mono = {'family': 'monospace'}
+        ax.text(label_position, rect.get_y(), labels[i], ha=align, va='bottom', rotation=0, color=color, fontdict=mono)
+
+def _plot_regions(dataset):
+    """
+    plot regions, sorted by latency criticality
+    """
+    relevant = sorted([x for x in dataset if x['Slowdown'] > 0 and x['Slowdown'] <= 5], key=lambda x: x['Slowdown'])
+    x = np.arange(len(relevant))
+
+    correlations = list(map(lambda x: x['Slowdown'], relevant))
+    ticks = list(map(lambda x: "<%s,%s>" % (x['Start'][:-2], x['End'][:-2]), relevant))
+    fig, ax = plt.subplots()
+    rects = ax.barh(x, correlations, align="center", tick_label="")
+    autolabel(rects, ax, ticks)
+
+    plt.tight_layout()
+    plt.savefig("normalized-slowdown.pdf")
+    plt.close()
+
+
+def duration_to_string(d):
+    return str(d['Start']) + "->" + str(d['Stop'])
+
+def analyse(df1, df2, config, export=False):
+    data1 = df1['EndToEnd_D']
+    data2 = df2['EndToEnd_D']
+    print("#1 e2e: " + str(np.mean(data1)) + " +- " + str(np.std(data1)))
+    print("#2 e2e: " + str(np.mean(data2)) + " +- " + str(np.std(data2)))
+
+    normalize = np.mean(data1) / np.mean(data2)
+
+    dataset = []
+
+    for d in latency.get_durations_2([df1, df2], config):
+        local1 = df1[d['Stop']] - df1[d['Start']]
+        local2 = df2[d['Stop']] - df2[d['Start']]
+
+        # too short -> ignore
+        if np.mean(local1) == 0 or np.mean(local2) == 0:
+            continue
+
+        slowdown = np.mean(local1) / np.mean(local2)
+
+        #print(duration_to_string(d)+" "+str(lf) + " / "+str(frac)+ " = " + str(lf/frac))
+        #print("\"<"+d['Start']+","+d['Stop']+">\" "+str(slowdown/normalize)+"   "+str(slowdown)+" "+str(normalize))
+        dataset += [{ 'Start': d['Start'], 'End': d['Stop'], 'Slowdown': slowdown }]
+
+    _plot_regions(dataset)
+
+

xlap/analyse/latency.py

@@ -204,6 +204,10 @@ def get_durations(df, config):
                 hdb += [{'Start': str(event1), 'Stop': str(event2), 'Source': 'cfa'}]
     return hdb
 
+def get_durations_2(dfs, config):
+    df = dfs[0] + dfs[1]
+    return get_durations(df, config)
+
 def analyse(df, config):
     hb = []
 

xlap/command_line.py

@@ -4,11 +4,13 @@ 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
+import xlap.analyse.e2e as e2e
 
 tasks = {
     "jitter": None,
     "latency": None,
     "difference": None,
+    "e2e": None,
     "capture": None
 }
 
@@ -43,18 +45,29 @@ def main():
                     f.write("\n")
             else:
                 print(output)
+        elif command == "e2e":
+            khz1 = 2000000
+            khz2 = 3000000
+            path="../publications/nsdi-18/eval/20180419_energy/"
+            df_data1 = evaluate(path+"sender-"+str(khz1)+".csv", path+"receiver-"+str(khz1)+".csv", config=config, kind=0)
+            df_data2 = evaluate(path+"sender-"+str(khz2)+".csv", path+"receiver-"+str(khz2)+".csv", config=config, kind=0)
+            e2e.analyse(df_data1, df_data2, config)
         elif command == "latency":
             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)
+            path="../publications/nsdi-18/eval/20180420_"
+            khz = 3000000
+            df_data1 = evaluate(path+"base1/sender-"+str(khz)+".csv", path+"base1/receiver-"+str(khz)+".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)
+            #df_data2 = evaluate("../prrt/out/s+send.csv", "../prrt/out/r+send.csv", config=config, kind=0)
+            df_data2 = evaluate(path+"base2/sender-"+str(khz)+".csv", path+"base2/receiver-"+str(khz)+".csv", config=config, kind=0)
+            #df_data2 = evaluate(path+"changed/sender-"+str(khz)+".csv", path+"changed/receiver-"+str(khz)+".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)