--- /dev/null
+#!/usr/bin/env python
+
+###############################################################################
+#
+# CCNX benchmark
+# Copyright (C) 2014 INRIA
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+#
+# Author: Alina Quereilhac <alina.quereilhac@inria.fr>
+#
+###############################################################################
+
+#
+# This library contains functions to parse (CCNx) ccnd logs.
+#
+# Results from experiments must be stored in a directory
+# named with the experiment run id.
+# ccnd logs are stored in .log files in a subdirectory per node.
+# The following diagram exemplifies the experiment result directory
+# structure (nidi is the unique identifier assigned to node i):
+#
+# run_id
+# \ nid1
+# \ nid2.log
+# \ nid2
+# \ nid1.log
+# \ nid3
+# \ nid3.log
+#
+
+import collections
+import functools
+import networkx
+import os
+import pickle
+import tempfile
+
+from nepi.util.timefuncs import compute_delay_ms
+from nepi.util.statfuncs import compute_mean
+import nepi.data.processing.ping.parser as ping_parser
+
+def is_control(content_name):
+ return content_name.startswith("ccnx:/%C1") or \
+ content_name.startswith("ccnx:/ccnx") or \
+ content_name.startswith("ccnx:/...")
+
+
+def parse_file(filename):
+ """ Parses message information from ccnd log files
+
+ filename: path to ccndlog file
+
+ """
+
+ faces = dict()
+ sep = " "
+
+ f = open(filename, "r")
+
+ data = []
+
+ for line in f:
+ cols = line.strip().split(sep)
+
+ # CCN_PEEK
+ # MESSAGE interest_from
+ # 1374181938.808523 ccnd[9245]: debug.4352 interest_from 6 ccnx:/test/bunny.ts (23 bytes,sim=0CDCC1D7)
+ #
+ # MESSAGE interest_to
+ # 1374181938.812750 ccnd[9245]: debug.3502 interest_to 5 ccnx:/test/bunny.ts (39 bytes,i=2844,sim=0CDCC1D7)
+ #
+ # MESSAGE CONTENT FROM
+ # 1374181938.868682 ccnd[9245]: debug.4643 content_from 5 ccnx:/test/bunny.ts/%FD%05%1E%85%8FVw/%00/%9E%3D%01%D9%3Cn%95%2BvZ%8
+ #
+ # MESSAGE CONTENT_TO
+ # 1374181938.868772 ccnd[9245]: debug.1619 content_to 6 ccnx:/test/bunny.ts/%FD%05%1E%85%8FVw/%00/%9E%3D%01%D9%3Cn%95%2BvZ%8
+ #
+ # 1375596708.222304 ccnd[9758]: debug.3692 interest_expiry ccnx:/test/bunny.ts/%FD%05%1E%86%B1GS/%00%0A%F7 (44 bytes,c=0:1,i=2819,sim=49FA8048)
+
+ # External face creation
+ # 1374181452.965961 ccnd[9245]: accepted datagram client id=5 (flags=0x40012) 204.85.191.10 port 9695
+
+ if line.find("accepted datagram client") > -1:
+ face_id = (cols[5]).replace("id=",'')
+ ip = cols[7]
+ port = cols[9]
+ faces[face_id] = (ip, port)
+ continue
+
+ # 1374181452.985296 ccnd[9245]: releasing face id 4 (slot 4)
+ if line.find("releasing face id") > -1:
+ face_id = cols[5]
+ if face_id in faces:
+ del faces[face_id]
+ continue
+
+ if len(cols) < 6:
+ continue
+
+ timestamp = cols[0]
+ message_type = cols[3]
+
+ if message_type not in ["interest_from", "interest_to", "content_from",
+ "content_to", "interest_dupnonce", "interest_expiry"]:
+ continue
+
+ face_id = cols[4]
+ content_name = cols[5]
+
+ # Interest Nonce ? -> 412A74-0844-0008-50AA-F6EAD4
+ nonce = ""
+ if message_type in ["interest_from", "interest_to", "interest_dupnonce"]:
+ last = cols[-1]
+ if len(last.split("-")) == 5:
+ nonce = last
+
+ try:
+ size = int((cols[6]).replace('(',''))
+ except:
+ print "interest_expiry without face id!", line
+ continue
+
+ # If no external IP address was identified for this face
+ # asume it is a local face
+ hostip = "localhost"
+
+ if face_id in faces:
+ hostip, port = faces[face_id]
+
+ data.append((content_name, timestamp, message_type, hostip, face_id,
+ size, nonce, line))
+
+ f.close()
+
+ return data
+
+def dump_content_history(content_history):
+ f = tempfile.NamedTemporaryFile(delete=False)
+ pickle.dump(content_history, f)
+ f.close()
+ return f.name
+
+def load_content_history(fname):
+ f = open(fname, "r")
+ content_history = pickle.load(f)
+ f.close()
+
+ os.remove(fname)
+ return content_history
+
+def annotate_cn_node(graph, nid, ips2nid, data, content_history):
+ for (content_name, timestamp, message_type, hostip, face_id,
+ size, nonce, line) in data:
+
+ # Ignore local messages for the time being.
+ # They could later be used to calculate the processing times
+ # of messages.
+ if peer == "localhost":
+ return
+
+ # Ignore control messages for the time being
+ if is_control(content_name):
+ return
+
+ if message_type == "interest_from" and \
+ peer == "localhost":
+ graph.node[nid]["ccn_consumer"] = True
+ elif message_type == "content_from" and \
+ peer == "localhost":
+ graph.node[nid]["ccn_producer"] = True
+
+ # remove digest
+ if message_type in ["content_from", "content_to"]:
+ content_name = "/".join(content_name.split("/")[:-1])
+
+ if content_name not in content_history:
+ content_history[content_name] = list()
+
+ peernid = ips2nid[peer]
+ add_edge(graph, nid, peernid)
+
+ content_history[content_name].append((timestamp, message_type, nid,
+ peernid, nonce, size, line))
+
+def annotate_cn_graph(logs_dir, graph, parse_ping_logs = False):
+ """ Adds CCN content history for each node in the topology graph.
+
+ """
+ # Make a copy of the graph to ensure integrity
+ graph = graph.copy()
+
+ ips2nids = dict()
+
+ for nid in graph.nodes():
+ ips = graph.node[nid]["ips"]
+ for ip in ips:
+ ips2nids[ip] = nid
+
+ # Now walk through the ccnd logs...
+ for dirpath, dnames, fnames in os.walk(logs_dir):
+ # continue if we are not at the leaf level (if there are subdirectories)
+ if dnames:
+ continue
+
+ # Each dirpath correspond to a different node
+ nid = os.path.basename(dirpath)
+
+ content_history = dict()
+
+ for fname in fnames:
+ if fname.endswith(".log"):
+ filename = os.path.join(dirpath, fname)
+ data = parse_file(filename)
+ annotate_cn_node(graph, nid, ips2nids, data, content_history)
+
+ # Avoid storing everything in memory, instead dump to a file
+ # and reference the file
+ fname = dump_content_history(content_history)
+ graph.node[nid]["history"] = fname
+
+ if parse_ping_logs:
+ ping_parser.annotate_cn_graph(logs_dir, graph)
+
+ return graph
+
+def ccn_producers(graph):
+ """ Returns the nodes that are content providers """
+ return [nid for nid in graph.nodes() \
+ if graph.node[nid].get("ccn_producer")]
+
+def ccn_consumers(graph):
+ """ Returns the nodes that are content consumers """
+ return [nid for nid in graph.nodes() \
+ if graph.node[nid].get("ccn_consumer")]
+
+def process_content_history(graph):
+ """ Compute CCN message counts and aggregates content historical
+ information in the content_names dictionary
+
+ """
+
+ ## Assume single source
+ source = ccn_consumers(graph)[0]
+
+ interest_expiry_count = 0
+ interest_dupnonce_count = 0
+ interest_count = 0
+ content_count = 0
+ content_names = dict()
+
+ # Collect information about exchanged messages by content name and
+ # link delay info.
+ for nid in graph.nodes():
+ # Load the data collected from the node's ccnd log
+ fname = graph.node[nid]["history"]
+ history = load_content_history(fname)
+
+ for content_name in history.keys():
+ hist = history[content_name]
+
+ for (timestamp, message_type, nid1, nid2, nonce, size, line) in hist:
+ if message_type in ["content_from", "content_to"]:
+ # The first Interest sent will not have a version or chunk number.
+ # The first Content sent back in reply, will end in /=00 or /%00.
+ # Make sure to map the first Content to the first Interest.
+ if content_name.endswith("/=00"):
+ content_name = "/".join(content_name.split("/")[0:-2])
+
+ # Add content name to dictionary
+ if content_name not in content_names:
+ content_names[content_name] = dict()
+ content_names[content_name]["interest"] = dict()
+ content_names[content_name]["content"] = list()
+
+ # Classify interests by replica
+ if message_type in ["interest_from"] and \
+ nonce not in content_names[content_name]["interest"]:
+ content_names[content_name]["interest"][nonce] = list()
+
+ # Add consumer history
+ if nid == source:
+ if message_type in ["interest_to", "content_from"]:
+ # content name history as seen by the source
+ if "consumer_history" not in content_names[content_name]:
+ content_names[content_name]["consumer_history"] = list()
+
+ content_names[content_name]["consumer_history"].append(
+ (timestamp, message_type))
+
+ # Add messages per content name and cumulate totals by message type
+ if message_type == "interest_dupnonce":
+ interest_dupnonce_count += 1
+ elif message_type == "interest_expiry":
+ interest_expiry_count += 1
+ elif message_type == "interest_from":
+ interest_count += 1
+ # Append to interest history of the content name
+ content_names[content_name]["interest"][nonce].append(
+ (timestamp, nid2, nid1))
+ elif message_type == "content_from":
+ content_count += 1
+ # Append to content history of the content name
+ content_names[content_name]["content"].append((timestamp, nid2, nid1))
+ else:
+ continue
+ del hist
+ del history
+
+ # Compute the time elapsed between the time an interest is sent
+ # in the consumer node and when the content is received back
+ for content_name in content_names.keys():
+ # order content and interest messages by timestamp
+ content_names[content_name]["content"] = sorted(
+ content_names[content_name]["content"])
+
+ for nonce, timestamps in content_names[content_name][
+ "interest"].iteritems():
+ content_names[content_name]["interest"][nonce] = sorted(
+ timestamps)
+
+ history = sorted(content_names[content_name]["consumer_history"])
+ content_names[content_name]["consumer_history"] = history
+
+ # compute the rtt time of the message
+ rtt = None
+ waiting_content = False
+ interest_timestamp = None
+ content_timestamp = None
+
+ for (timestamp, message_type) in history:
+ if not waiting_content and message_type == "interest_to":
+ waiting_content = True
+ interest_timestamp = timestamp
+ continue
+
+ if waiting_content and message_type == "content_from":
+ content_timestamp = timestamp
+ break
+
+ # If we can't determine who sent the interest, discard it
+ rtt = -1
+ if interest_timestamp and content_timestamp:
+ rtt = compute_delay_ms(content_timestamp, interest_timestamp)
+
+ content_names[content_name]["rtt"] = rtt
+ content_names[content_name]["lapse"] = (interest_timestamp, content_timestamp)
+
+ return (content_names,
+ interest_expiry_count,
+ interest_dupnonce_count,
+ interest_count,
+ content_count)
+
+def process_content_history_logs(logs_dir, graph):
+ """ Parse CCN logs and aggregate content history information in graph.
+ Returns annotated graph and message countn and content names history.
+
+ """
+ ## Process logs and analyse data
+ try:
+ graph = annotate_cn_graph(logs_dir, graph,
+ parse_ping_logs = True)
+ except:
+ print "Skipping: Error parsing ccnd logs", logs_dir
+ raise
+
+ source = consumers(graph)[0]
+ target = producers(graph)[0]
+
+ # Process the data from the ccnd logs, but do not re compute
+ # the link delay.
+ try:
+ (graph,
+ content_names,
+ interest_expiry_count,
+ interest_dupnonce_count,
+ interest_count,
+ content_count) = process_content_history(graph)
+ except:
+ print "Skipping: Error processing ccn data", logs_dir
+ raise
+
+ return (graph,
+ content_names,
+ interest_expiry_count,
+ interest_dupnonce_count,
+ interest_count,
+ content_count)
git://git.onelab.eu / nepi.git / blobdiff