commit 4f4bc10fba85adc877481e3fb4445847a208da12 Author: Mike Perry mikeperry-git@fscked.org Date: Thu Dec 1 20:09:25 2011 -0800

Implement per-class setpoints.

Governed by consensus param (off by default). --- NetworkScanners/BwAuthority/README.spec.txt | 5 ++ NetworkScanners/BwAuthority/aggregate.py | 65 ++++++++++++++++++++++----- 2 files changed, 58 insertions(+), 12 deletions(-)

diff --git a/NetworkScanners/BwAuthority/README.spec.txt b/NetworkScanners/BwAuthority/README.spec.txt index 81468b5..acce106 100644 --- a/NetworkScanners/BwAuthority/README.spec.txt +++ b/NetworkScanners/BwAuthority/README.spec.txt @@ -528,6 +528,11 @@ accumulated in pid_error_sum as opposed to the consensus value itself.

+ "bwauthbyclass=1" + If present, computes F_avg (see Section 3.1) for each class + of Guard, Middle, Exit, and Guard+Exit nodes, and uses these + flag-specific averages to compute pid_error. + "bwauthkp=N" Sets K_p to N/10000.0. If absent, K_p=1.0.

diff --git a/NetworkScanners/BwAuthority/aggregate.py b/NetworkScanners/BwAuthority/aggregate.py index 6283f3f..120de07 100755 --- a/NetworkScanners/BwAuthority/aggregate.py +++ b/NetworkScanners/BwAuthority/aggregate.py @@ -92,6 +92,7 @@ class Node: self.new_bw = None self.change = None self.use_bw = -1 + self.flags = ""

# measurement vars from bwauth lines self.measured_at = 0 @@ -132,6 +133,16 @@ class Node: + kd*self.use_bw*self.d_error_dt() return self.pid_bw

+ def node_class(self): + if "Guard" in self.flags and "Exit" in self.flags: + return "Guard+Exit" + elif "Guard" in self.flags: + return "Guard" + elif "Exit" in self.flags: + return "Exit" + else: + return "Middle" + # Time-weighted sum of error per unit of time (measurement sample) def integral_error(self): if self.prev_error == 0: @@ -222,6 +233,7 @@ class ConsensusJunk: def __init__(self, c): cs_bytes = c.sendAndRecv("GETINFO dir/status-vote/current/consensus\r\n")[0][2] self.bwauth_pid_control = True + self.group_by_class = False self.use_circ_fails = False self.use_best_ratio = True self.use_desc_bw = True @@ -246,6 +258,9 @@ class ConsensusJunk: elif p == "bwauthbestratio=0": self.use_best_ratio = False plog("INFO", "Choosing larger of sbw vs fbw") + elif p == "bwauthbyclass=1": + self.group_by_class = True + plog("INFO", "Grouping nodes by flag-class") elif p.startswith("bwauthkp="): self.K_p = int(p.split("=")[1])/10000.0 plog("INFO", "Got K_p=%f from consensus." % self.K_p) @@ -439,23 +454,45 @@ def main(argv): for n in nodes.itervalues(): n.circ_fail_rate = 0.0

+ for idhex in nodes.iterkeys(): + if idhex in prev_consensus: + nodes[idhex].flags = prev_consensus[idhex].flags + + true_filt_avg = {} + true_strm_avg = {} + if cs_junk.bwauth_pid_control: # Penalize nodes for circuit failure: it indicates CPU pressure # TODO: Potentially penalize for stream failure, if we run into # socket exhaustion issues.. plog("INFO", "PID control enabled") - true_filt_avg = sum(map(lambda n: n.filt_bw*(1.0-n.circ_fail_rate), - nodes.itervalues()))/float(len(nodes)) - true_strm_avg = sum(map(lambda n: n.strm_bw*(1.0-n.circ_fail_rate), - nodes.itervalues()))/float(len(nodes)) + + if cs_junk.group_by_class: + for c in ["Guard+Exit", "Guard", "Exit", "Middle"]: + c_nodes = filter(lambda n: n.node_class() == c, nodes.itervalues()) + true_filt_avg[c] = sum(map(lambda n: n.filt_bw*(1.0-n.circ_fail_rate), + c_nodes))/float(len(c_nodes)) + true_strm_avg[c] = sum(map(lambda n: n.strm_bw*(1.0-n.circ_fail_rate), + c_nodes))/float(len(c_nodes)) + plog("INFO", "Network true_filt_avg["+c+"]: "+str(true_filt_avg[c])) + else: + filt_avg = sum(map(lambda n: n.filt_bw*(1.0-n.circ_fail_rate), + nodes.itervalues()))/float(len(nodes)) + strm_avg = sum(map(lambda n: n.strm_bw*(1.0-n.circ_fail_rate), + nodes.itervalues()))/float(len(nodes)) + for c in ["Guard+Exit", "Guard", "Exit", "Middle"]: + true_filt_avg[c] = filt_avg + true_strm_avg[c] = strm_avg else: plog("INFO", "PID control disabled") - true_filt_avg = sum(map(lambda n: n.filt_bw, - nodes.itervalues()))/float(len(nodes)) - true_strm_avg = sum(map(lambda n: n.strm_bw, + filt_avg = sum(map(lambda n: n.filt_bw*(1.0-n.circ_fail_rate), + nodes.itervalues()))/float(len(nodes)) + strm_avg = sum(map(lambda n: n.strm_bw*(1.0-n.circ_fail_rate), nodes.itervalues()))/float(len(nodes)) + for c in ["Guard+Exit", "Guard", "Exit", "Middle"]: + true_filt_avg[c] = filt_avg + true_strm_avg[c] = strm_avg

- plog("DEBUG", "Network true_filt_avg: "+str(true_filt_avg))

prev_votes = None if cs_junk.bwauth_pid_control: @@ -489,8 +526,8 @@ def main(argv):

tot_net_bw = 0 for n in nodes.itervalues(): - n.fbw_ratio = n.filt_bw/true_filt_avg - n.sbw_ratio = n.strm_bw/true_strm_avg + n.fbw_ratio = n.filt_bw/true_filt_avg[n.node_class()] + n.sbw_ratio = n.strm_bw/true_strm_avg[n.node_clasS()]

if cs_junk.bwauth_pid_control: if cs_junk.use_desc_bw: @@ -500,9 +537,13 @@ def main(argv):

# Penalize nodes for circ failure rate if cs_junk.use_best_ratio and n.sbw_ratio > n.fbw_ratio: - n.pid_error = (n.strm_bw*(1.0-n.circ_fail_rate) - true_strm_avg)/true_strm_avg + n.pid_error = (n.strm_bw*(1.0-n.circ_fail_rate) - + true_strm_avg[n.node_class()]) \ + / true_strm_avg[n.node_class()] else: - n.pid_error = (n.filt_bw*(1.0-n.circ_fail_rate) - true_filt_avg)/true_filt_avg + n.pid_error = (n.filt_bw*(1.0-n.circ_fail_rate) - + true_filt_avg[n.node_class()]) \ + / true_filt_avg[n.node_class()]

if n.idhex in prev_votes.vote_map: # If there is a new sample, let's use it for all but guards