more

.gitignore

@@ -0,0 +1,3 @@
+**/.DS_Store
+rln-delay-simulations/.idea/*
+**/__pycache__

rln-delay-simulations/README.md

@@ -6,6 +6,7 @@ This folder contains a `shadow` configuration to simulate `1000` `nwaku` nodes i
 * Focused on measuring message propagation delays. Each message that is sent, encodes the timestamp when it was created.
 * Requires significant resources to run (tested with 256 GB RAM)
 * See simulation parameters: latency, bandwidth, amount of nodes, amount of publishers.
+* Note that due to TCP flow control, when using big messages the first ones to arrive will show a higher delay. Filter them out to not bias the measurements.
 
 ## How to run
 
@@ -27,7 +28,10 @@ grep -nr 'ended_simulation' shadow.data | wc -l
 # expected: 1000 (simulation finished ok in all nodes)
 
 grep -nr 'tx_msg' shadow.data | wc -l
-# expected: 15 (total of published messages)
+# expected: 10 (total of published messages)
+
+grep -nr 'rx_msg' shadow.data | wc -l
+# expected: 9990 (total rx messages)
 ```
 
 Get metrics:

rln-delay-simulations/analyze.py

@@ -1,23 +1,26 @@
 import numpy as np
 import sys
+def load(file_name, field):
+    latencies = []
+    with open(file_name, "r") as file:
+        for line in file.readlines():
+            if field in line:
+                # first message bias the latency due to tcp flow control
+                if "seq=" in line:
+                    seq = int(line.strip().split("seq=")[1].split(" ")[0])
+                    if seq in [0]:
+                        continue
+                x = line.strip().split(field)[1].split(" ")[0]
+                latencies.append(int(x))
+    return np.array(latencies)
 
-file = sys.argv[1]
-field = sys.argv[2]
-print("Data file:", file, "field:", field)
 
-latencies = []
-with open(file, "r") as file:
-    for line in file.readlines():
-        if field in line:
-            # first message bias the latency due to tcp flow control
-            if "seq=" in line:
-                seq = int(line.strip().split("seq=")[1].split(" ")[0])
-                if seq in [0]:
-                    continue
-            x = line.strip().split(field)[1].split(" ")[0]
-            latencies.append(int(x))
+if __name__ == "__main__":
+    file = sys.argv[1]
+    field = sys.argv[2]
+    array = load(file, field)
 
-array = np.array(latencies)
-print(f"number_samples={array.size}")
-print(f"Percentiles. P75={np.percentile(array, 75)} P95={np.percentile(array, 95)}")
-print(f"Statistics. mean={np.mean(array)} max={array.max()} min={array.min()}")
+    print("Data file:", file, "field:", field)
+    print(f"number_samples={array.size}")
+    print(f"Percentiles. P75={np.percentile(array, 75)} P95={np.percentile(array, 95)}")
+    print(f"Statistics. mean={np.mean(array)} max={array.max()} min={array.min()}")

rln-delay-simulations/plot.py

@@ -1,55 +1,33 @@
 import matplotlib.pyplot as plt
 import scienceplots
 import numpy as np
-import math
-import random
+from analyze import load
 
-def load(file):
-    field = "arrival_diff="
-    latencies = []
-    with open(file, "r") as file:
-        for line in file.readlines():
-            if field in line:
-                x = line.strip().split(field)[1].split(" ")[0]
-                latencies.append(int(x))
-    return np.array(latencies)
+latencies = [load("raw/latency_10kb.txt", "arrival_diff="),
+             load("raw/latency_100kb.txt", "arrival_diff="),
+             load("raw/latency_500kb.txt", "arrival_diff=")]
 
-latencies = [load("latency_10kb.txt"), load("latency_100kb.txt"), load("latency_500kb.txt")]
+print(latencies)
 
 with plt.style.context(['science', 'ieee']):
     fig, ax = plt.subplots()
-
-    labels = []
-
     bp = ax.boxplot(latencies,notch=True, vert=True, patch_artist=True,
                     showfliers=True, whis=100000000000)
 
     for patch, color in zip(bp['boxes'], ['red', 'blue', 'green']):
         patch.set_facecolor(color)
 
-    ax.set(title="Message rate (msg/s)", xlabel='Scenario', ylabel='Message propagation time (ms)')
-
+    ax.set(title="Message latencies distribution\nD=6 nodes=1000 samples="+str(latencies[1].size), xlabel='Scenario', ylabel='Message propagation time (ms)')
     ax.grid(linestyle='-')
     my_legend = []
     for msg_size in [10, 100, 500]:
         my_legend.append("Message size: " + str(msg_size) + " kB")
     ax.legend([bp["boxes"][i] for i in range(len(my_legend))], my_legend, loc='upper left', fontsize=5)
 
-    ax.autoscale(tight=False)
-
-#means = [i.mean(axis=0) for i in q]
-#stds =[i.std(axis=0) for i in q]
-#per_95 = [np.percentile(i, 95) for i in q]
-#for i, line in enumerate(bp['medians']):
-#    x, y = line.get_xydata()[1]
-#    text = r' $ \mu =$ ' + '{:.2f}\n'.format(means[i]) + r' $ p_{95} = $ ' + '{:.2f}'.format(per_95[i])
-#    ax.annotate(text, xy=(x+0.1, y), fontsize=6)
-#
-#
-
 for i, line in enumerate(bp['medians']):
     x, y = line.get_xydata()[1]
     text = r' $ \mu =$ ' + '{:.0f} ms\n'.format(latencies[i].mean(axis=0)) + r' $ p_{95} = $ ' + '{:.0f} ms'.format(np.percentile(latencies[i], 95))
     ax.annotate(text, xy=(x + 0.1, y), fontsize=6)
 
+fig.set_size_inches(4, 3)
 fig.savefig('plot.jpg', dpi=600)