[tor-commits] [metrics-tasks/master] Add graph with single relay probabilities (#1854).

[karsten at torproject.org]

commit 7d94e96a9989342b7dd3a78705cfd41a9611853f
Author: Karsten Loesing <karsten.loesing at gmx.net>
Date:   Tue Nov 27 10:41:14 2012 -0500

    Add graph with single relay probabilities (#1854).
---
 task-1854/plot-entropy.R |   29 ++++++++++++++++++++++++++---
 task-1854/pylinf.py      |   13 ++++++++-----
 2 files changed, 34 insertions(+), 8 deletions(-)

diff --git a/task-1854/plot-entropy.R b/task-1854/plot-entropy.R
index 5d30583..cc963e4 100644
--- a/task-1854/plot-entropy.R
+++ b/task-1854/plot-entropy.R
@@ -3,7 +3,7 @@ library(reshape)
 library(scales)
 
 p <- read.csv("prob-extracted.csv", header = FALSE,
-  col.names = c("validafter", "minadvbw", "advbw", "cumprob"),
+  col.names = c("validafter", "minadvbw", "advbw", "cumprob", "prob"),
   stringsAsFactor = FALSE)
 p <- p[p$minadvbw >= 20480, ]
 c <- data.frame(x = p$advbw, y = p$cumprob,
@@ -13,14 +13,37 @@ geom_line() +
 scale_x_log10(name = "\nAdvertised bandwidth in B/s (log scale)") +
 scale_y_continuous(name = "Cumulative probability\n") +
 scale_colour_hue(name = "Adv. bw. cutoff in B/s") +
-opts(legend.position = "top")
+opts(title = paste("Consensus with valid-after time", max(p$validafter)),
+  legend.position = "top")
 
 ggplot(c, aes(x = x, y = y, colour = colour)) +
 geom_line() +
 scale_x_log10(name = "\nAdvertised bandwidth in B/s (log scale)") +
 scale_y_log10(name = "Cumulative probability (log scale)\n") +
 scale_colour_hue(name = "Adv. bw. cutoff in B/s") +
-opts(legend.position = "top")
+opts(title = paste("Consensus with valid-after time", max(p$validafter)),
+  legend.position = "top")
+
+p <- p[p$minadvbw == 20480 | p$minadvbw == 1048576, ]
+c <- data.frame(x = p$advbw, y = p$prob,
+  colour = as.factor(p$minadvbw))
+ggplot(c, aes(x = x, y = y, colour = colour)) +
+geom_point(alpha = 0.25) +
+scale_x_log10(name = "\nAdvertised bandwidth in B/s (log scale)") +
+scale_y_continuous(name = paste("Single relay probability, *not*",
+  "probability distribution function\n")) +
+scale_colour_hue(name = "Advertised bandwidth cutoff in B/s") +
+opts(title = paste("Consensus with valid-after time", max(p$validafter)),
+  legend.position = "top")
+
+ggplot(c, aes(x = x, y = y, colour = colour)) +
+geom_point(alpha = 0.25) +
+scale_x_log10(name = "\nAdvertised bandwidth in B/s (log scale)") +
+scale_y_log10(name = paste("Single relay probability, *not*",
+  "probability distribution function (log scale)\n")) +
+scale_colour_hue(name = "Advertised bandwidth cutoff in B/s") +
+opts(title = paste("Consensus with valid-after time", max(p$validafter)),
+  legend.position = "top")
 
 e <- read.csv("linf-extracted.csv", header = FALSE,
   col.names = c("validafter", "min_adv_bw", "relays", "linf",
diff --git a/task-1854/pylinf.py b/task-1854/pylinf.py
index 9885de5..15833e6 100644
--- a/task-1854/pylinf.py
+++ b/task-1854/pylinf.py
@@ -168,8 +168,8 @@ def run(data):
                                       str(len(routers)-omitted_routers),
                                       str(max(prob_diff))]))
 
-        while len(cutoffs) > 0 and min_adv_bw > cutoffs[0]:
-            cumulated_prob = 0.0
+        while len(cutoffs) > 0 and min_adv_bw >= cutoffs[0]:
+            prev_prob, cumulated_prob = 0.0, 0.0
             prev_advertised_bw = 0
             for router in routers:
                 if router.advertised_bw > cutoffs[0] and \
@@ -177,13 +177,16 @@ def run(data):
                     prob_string.append(','.join([valid_after,
                                                  str(cutoffs[0]),
                                                  str(prev_advertised_bw),
-                                                 str(cumulated_prob)]))
+                                                 str(cumulated_prob),
+                                                 str(prev_prob)]))
                 prev_advertised_bw = router.advertised_bw
-                cumulated_prob += float(router.bandwidth)/float(total_bw)
+                prev_prob = float(router.bandwidth)/float(total_bw)
+                cumulated_prob += prev_prob
             prob_string.append(','.join([valid_after,
                                          str(cutoffs[0]),
                                          str(prev_advertised_bw),
-                                         str(cumulated_prob)]))
+                                         str(cumulated_prob),
+                                         str(prev_prob)]))
             cutoffs.pop(0)
 
         # remove routers with min adv_bw