commit 18509bdf00646fda601581795c6f7d7800976a50 Author: Tomás Touceda chiiph@torproject.org Date: Thu May 10 19:41:09 2012 -0300

Add simple parser for bridge descriptors in python to work on ticket 2866 --- task-2866/parser.py | 313 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 313 insertions(+), 0 deletions(-)

diff --git a/task-2866/parser.py b/task-2866/parser.py new file mode 100644 index 0000000..260120b --- /dev/null +++ b/task-2866/parser.py @@ -0,0 +1,313 @@ +import os +import datetime +import sys + +from math import floor + +UNALLOC_POOL = "unallocated" + +class Bridge(object): + def __init__(self, fp, pool = "", params = {}): + object.__init__(self) + self._fp = fp + self._pool = pool + self._params = params + self._from = None + self._to = None + self._usage_per_country = {} + self._usage_per_day = {} + + def pprint(self): + print self._fp, self._pool, self._params + + # <40 bytes fp> <pool> <key>=<val> ... + @staticmethod + def fromLine(line): + fp = line[:40] + pool = "" + params = {} + + rest = line[41:] + next_spc = rest.find(" ") + if next_spc == -1: + pool = rest + return Bridge(fp, pool) + + pool = rest[:next_spc] + rest = rest[next_spc+1:] + + key = val = "" + next_spc = rest.find(" ") + while next_spc != -1: + key, val = Bridge.parseKeyVal(rest[:next_spc]) + params[key] = val + rest = rest[next_spc+1:] + next_spc = rest.find(" ") + + return Bridge(fp, pool, params) + + @staticmethod + def parseKeyVal(line): + [key, val] = line.split("=") + return key, val + +class BridgePool(object): + def __init__(self, path, bridges, year, month, day, hour, minute, seconds): + object.__init__(self) + self._path = path + self._date = (year, month, day, hour, minute, seconds) + self._bridges = bridges + self.gatherBridges() + + def pprint(self): + print "%s/%s/%s :: %s:%s:%s" % self._date + for b in self._bridges: + b.pprint() + + def gatherBridges(self): + f = open(self._path, "r") + contents = f.read().split("\n")[1:] + for line in contents: + if len(line.strip()) == 0: + continue + if line[:40] in self._bridges.keys(): + self._bridges[line[:40]]._to = self._date + else: + b = Bridge.fromLine(line) + b._from = self._date + self._bridges[b._fp] = b + +def listHas(l, comp): + for e in l: + if e.startswith(comp): + return True + return False + +class BridgeUsageAnalyzer(object): + def __init__(self, fromPath = ".", bridges = {}): + object.__init__(self) + self._bridges = bridges + self._root = fromPath + self.gatherUsage() + + def gatherUsage(self): + print "Walking dirs to gather bridge usage (you might want to go do something else for a while)..." + for root, dirs, files in os.walk(self._root): + comps = root.split(os.sep) + # bridge-descriptors-<year>-<month>/fp[0]/fp[1]/fp + if not listHas(comps, "bridge-descriptors-"): + continue + if not listHas(comps, "extra-info"): + continue + print root + + for f in files: + f_value = os.path.join(root, f) + try: + bfp, ips, day = self.parseExtraInfo(f_value) + if not bfp in self._bridges.keys(): + #print "WARNING: %s is not in the bridge pool" % (f,) + continue + daily_usage = 0 + for c in ips.keys(): + if not c in self._bridges[bfp]._usage_per_country.keys(): + self._bridges[bfp]._usage_per_country[c] = 0 + self._bridges[bfp]._usage_per_country[c] += ips[c] + daily_usage += ips[c] + if not day in self._bridges[bfp]._usage_per_day.keys(): + self._bridges[bfp]._usage_per_day[day] = 0 + self._bridges[bfp]._usage_per_day[day] += daily_usage + except KeyboardInterrupt, e: + sys.exit(1) + except Exception, e: + print e + print "ERROR: Something was wrong while processing %s" % (f_value,) + + def parseExtraInfo(self, path): + contents = open(path, "r").read() + bfp = "" + ips = {} + day = "" + for l in contents.split("\n"): + if l.startswith("extra-info Unnamed"): + bfp = l.split(" ")[-1].lower() + continue + if l.startswith("published"): + day = l.split(" ")[1] + if l.startswith("bridge-ips"): + bridgeips = l.split(",")[1:] + try: + for ci in bridgeips: + parts = ci.split("=") + ips[parts[0]] = int(parts[1]) + except: + break + + return bfp, ips, day + + # Question: is there any correlation between the usage by country + # and the pool the bridge is from? + def analyzeUsage(self): + alloc_users = 0 + unalloc_users = 0 + + alloc_usage = {} + unalloc_usage = {} + for bfp in self._bridges.keys(): + acc = 0 + for c in self._bridges[bfp]._usage_per_country.keys(): + acc = self._bridges[bfp]._usage_per_country[c] + if self._bridges[bfp]._pool == UNALLOC_POOL: + unalloc_users += acc + if not c in unalloc_usage.keys(): + unalloc_usage[c] = 0 + unalloc_usage[c] += acc + else: + alloc_users += acc + if not c in alloc_usage.keys(): + alloc_usage[c] = 0 + alloc_usage[c] += acc + + print "Total allocated pool users: %d" % (alloc_users,) + print "Total unallocated pool users: %d" % (unalloc_users,) + + alloc_items = alloc_usage.items() + alloc_sorted_items = sorted(alloc_items, key=lambda use: use[1]) + print "Country usage for allocated pool:" + for country, count in alloc_sorted_items: + print " %s = %d" % (country, count) + + unalloc_items = unalloc_usage.items() + unalloc_sorted_items = sorted(unalloc_items, key=lambda use: use[1]) + print "Country usage for unallocated pool:" + for country, count in unalloc_sorted_items: + print " %s = %d" % (country, count) + + total_avg_unalloc = [] + total_avg_alloc = [] + for bfp in self._bridges.keys(): + avg = 0.0 + for day in self._bridges[bfp]._usage_per_day.keys(): + avg += self._bridges[bfp]._usage_per_day[day] + if avg == 0.0: + continue + avg = float(avg)/len(self._bridges[bfp]._usage_per_day.keys()) + if self._bridges[bfp]._pool == UNALLOC_POOL: + total_avg_unalloc.append(avg) + else: + total_avg_alloc.append(avg) + #print "Average for %s: %d" % (bfp, avg) + total_avg_unalloc.sort() + total_avg_alloc.sort() + + print "Unallocated pool:" + print " Average clients per day: %d" % (float(sum(total_avg_unalloc))/len(total_avg_unalloc),) + print " Median: %d" % (total_avg_unalloc[int(round(len(total_avg_unalloc)/2.0))],) + print " 75th percentile: %d" % (total_avg_unalloc[int(round(len(total_avg_unalloc)*0.75))],) + print " 90th percentile: %d" % (total_avg_unalloc[int(round(len(total_avg_unalloc)*0.9))],) + + print "Allocated pool:" + print " Average clients per day: %d" % (float(sum(total_avg_alloc))/len(total_avg_alloc),) + print " Median: %d" % (total_avg_alloc[int(round(len(total_avg_alloc)/2.0))],) + print " 75th percentile: %d" % (total_avg_alloc[int(round(len(total_avg_alloc)*0.75))],) + print " 90th percentile: %d" % (total_avg_alloc[int(round(len(total_avg_alloc)*0.9))],) + +class BridgePoolAnalyzer(object): + year = 0 + month = 1 + day = 2 + hour = 3 + minute = 4 + seconds = 5 + + def __init__(self, fromPath = "."): + object.__init__(self) + self._root = fromPath + self._pools = [] + self._bridges = {} + self.gatherBridgePools() + + def pprint(self): + for bp in self._pools: + bp.pprint() + + def gatherBridgePools(self): + print "Walking dirs to gather bridge pools (you might want to go do something else for a while)..." + for root, dirs, files in os.walk(self._root): + comps = root.split(os.sep) + if not listHas(comps,"bridge-pool-assignments"): + continue + print root + + for f in files: + f_value = os.path.join(root, f) + date = f.split("-") + try: + self._pools.append(BridgePool(f_value, self._bridges, + date[self.year], date[self.month], date[self.day], + date[self.hour], date[self.minute], date[self.seconds])) + except KeyboardInterrupt, e: + sys.exit(1) + except Exception, e: + print e + print "ERROR: Something was wrong while processing %s" % (f_value,) + + # Question: how much do bridges last when they are unassigned? + def bridgeAvailability(self): + print "Analyzing the data..." + availability_unalloc = {} + availability_else = {} + + for bfp in self._bridges.keys(): + b = self._bridges[bfp] + if b._to is None: + b._to = self._pools[-1]._date + if b._pool == UNALLOC_POOL: + availability_unalloc[b._fp] = (datetime.datetime(int(b._to[0]), int(b._to[1]), int(b._to[2]), + int(b._to[3]), int(b._to[4]), int(b._to[5])) - + datetime.datetime(int(b._from[0]), int(b._from[1]), int(b._from[2]), + int(b._from[3]), int(b._from[4]), int(b._from[5]))).total_seconds() / (60.0 * 60.0 * 24.0) + else: + availability_else[b._fp] = (datetime.datetime(int(b._to[0]), int(b._to[1]), int(b._to[2]), + int(b._to[3]), int(b._to[4]), int(b._to[5])) - + datetime.datetime(int(b._from[0]), int(b._from[1]), int(b._from[2]), + int(b._from[3]), int(b._from[4]), int(b._from[5]))).total_seconds() / (60.0 * 60.0 * 24.0) + + print "Allocated bridges:" + doall(availability_else) + + print "Unallocated bridges:" + doall(availability_unalloc) + +def calc_average(array): + avg = 0.0 + for k in array.keys(): + avg += float(array[k]) + avg /= len(array.keys()) + return avg + +def calc_median(values): + return values[int(floor(len(values)/2.0))] + +def calc_percentile(values, perc): + pos = int(round(perc/100.0 * float(len(values)))) + return values[pos] + +def doall(array): + print " Average availability:", calc_average(array) + values = array.values() + values.sort() + print " Median availability:", calc_median(values) + print " 75th percentile availability:", calc_percentile(values, 75) + print " 90th percentile availability:", calc_percentile(values, 90) + +if __name__ == "__main__": + path = "." + if len(sys.argv) > 1: + path = sys.argv[1] + bpa = BridgePoolAnalyzer(path) + #bpa.pprint() + bpa.bridgeAvailability() + + bua = BridgeUsageAnalyzer(path, bpa._bridges) + bua.analyzeUsage()