def upwalker_counter(g, meta, vf_g, arguments):
N = g.vcount()
progress = progressbar1.DummyProgressBar(end=10, width=15)
if arguments.progressbar:
progress = progressbar1.AnimatedProgressBar(end=len(meta), width=15)
# trace_up_map = {}
# random_up_map = {}
trace_up_counter = []
random_up_counter = []
for m in meta:
progress += 1
progress.show_progress()
trace = m[helpers.TRACE]
trace_dir = vft.trace_to_string(g, trace)
trace_up_count = trace_dir.count('U')
trace_up_counter.append(trace_up_count)
# trace_up_map[trace_up_count] = trace_up_map.get(trace_up_count, 0) +
# 1
s, t = trace[0], trace[-1]
s_idx, t_idx = vft.node_to_nodeid(g, s), vft.node_to_nodeid(g, t)
random_vf_route = helpers.random_route_walk(vf_g,
s_idx,
t_idx + N,
len(trace),
named=False,
weight_field='VFweight')
random_vf_route = vft.vf_route_converter(random_vf_route, N)
random_vf_dir = vft.trace_to_string(g, random_vf_route)
random_vf_count = random_vf_dir.count('U')
random_up_counter.append(random_vf_count)
# random_up_map[random_vf_count] = random_up_map.get(random_vf_count,
# 0) + 1
real_counter = collections.Counter(trace_up_counter)
real_up = ' '.join(
['%s: %s' % (k, real_counter[k]) for k in sorted(list(real_counter))])
random_counter = collections.Counter(random_up_counter)
random_up = ' '.join([
'%s: %s' % (k, random_counter[k]) for k in sorted(list(random_counter))
])
logger.info('')
logger.info('Real trace UP counter: %s' % real_up)
logger.info('Random vf trace up counter: %s' % random_up)
helpers.save_to_json(arguments.out, {
'REAL': dict(real_counter),
'RANDOM': dict(random_counter)
})
keys = sorted(set(list(real_counter) + list(random_counter)))
logger.info('IDX;REAL;RANDOM')
for k in keys:
logger.info('%s;%d;%d' % (k, real_counter[k], random_counter[k]))
def ba_generator(ba_graph, sh_paths, stretch, vf_g, progressbar=False):
vf_count = 0
trace_count = 0
lp_count = 0
progress = progressbar1.DummyProgressBar(end=10, width=15)
if progressbar:
progress = progressbar1.AnimatedProgressBar(end=len(sh_paths),
width=15)
for (s, t), shl in sh_paths:
progress += 1
progress.show_progress()
logger.debug('SH from %s to %s is %d' % (s, t, shl))
random_route = helpers.random_route_walk(ba_graph, s, t, shl + stretch)
logger.debug('Random route: %s' % random_route)
real_stretch = len(random_route) - shl
if real_stretch != stretch:
continue
trace_count += 1
is_vf = vft.is_valley_free(ba_graph,
random_route,
vfmode=vft.CLOSENESS)
logger.debug(
'Trace edge dir: %s' %
vft.trace_to_string(ba_graph, random_route, vfmode=vft.CLOSENESS))
logger.debug('Is VF: %s' % is_vf)
if is_vf:
is_lp = vft.is_local_preferenced(ba_graph,
random_route,
first_edge=True,
vfmode=vft.CLOSENESS,
vf_g=vf_g)
else:
is_lp = 0
logger.debug('Is LP: %s' % is_lp)
vf_count += int(is_vf)
lp_count += int(is_lp)
logger.info('Stretch %d' % stretch)
logger.info('Trace count: %d' % trace_count)
logger.info('VF count: %d' % vf_count)
logger.info('LP count: %d' % lp_count)
return (stretch, trace_count, vf_count, lp_count)
def filter(g, traceroutes):
results = list()
# remove traces with unknown nodes
traceroutes = vft.trace_in_vertex_id(g, traceroutes)
progress = progressbar1.AnimatedProgressBar(end=len(traceroutes), width=15)
for trace in traceroutes:
progress += 1
progress.show_progress()
if not vft.trace_exists(g, trace):
print 'BUG?'
continue
for x in range(0, g.vcount()):
g.vs[x]['traces'] = dict()
trace = tuple(trace)
s, t = trace[0], trace[-1]
sh_len = g.shortest_paths(s, t, mode=i.ALL)[0][0]
sh_len += 1 # igraph's hop count to node count
all_routes = helpers.dfs_mark(g, s, t, sh_len + 1)
# all_routes2 = helpers.dfs_simple(g, s, t, sh_len + 1, ())
# if set(all_routes) - set(all_routes2) != set(all_routes2) - set(all_routes):
# print 'AJAJAJ'
# print all_routes
# print '----------'
# print all_routes2
sh_routes = [x for x in all_routes if len(x) == sh_len]
all_vf_routes = [x for x in all_routes if vft.is_valley_free(g, x)]
prediction_set = set(sh_routes) | set(all_vf_routes)
result = [
trace,
len(trace),
sh_len,
len(sh_routes),
trace in sh_routes,
len(all_vf_routes),
trace in all_vf_routes,
len(all_routes),
trace in all_routes,
len(prediction_set),
trace in prediction_set,
vft.is_valley_free(g, trace),
# sh_routes, all_vf_routes, all_routes,
vft.trace_to_string(g, trace)
]
results.append(result)
print >> sys.stderr, (
'TRACE\tTRACE_LEN\tSH_LEN',
'\t#SH_ROUTE\tOK',
'\t#ALL_VF\tOK',
'\t#ALL_ROUTE\tOK',
'\t#PREDICTION_SET\tOK',
'\tIS_VF',
# '\tSH_ROUTES\tALL_VF_ROUTES\tALL_ROUTE',
'\tTRACE_STR')
for result in results:
result = [str(r) for r in result]
print >> sys.stderr, '\t'.join(result)
return results
def trace_dir(g, meta, vf_g, arguments):
N = g.vcount()
vf_sum_lp_w = sum([e['LPweight'] for e in vf_g.es])
out_instead_core = 0
core_instead_out = 0
in_customer = 0
in_customer_but_use_up = 0
progress = progressbar1.DummyProgressBar(end=10, width=15)
if arguments.progressbar:
progress = progressbar1.AnimatedProgressBar(end=len(meta), width=15)
for m in meta:
progress += 1
progress.show_progress()
trace = m[helpers.TRACE]
first_edge = [trace[0], trace[1]]
first_hop_type = vft.edge_dir(g, first_edge, vfmode=vft.CLOSENESS)
s_idx = vft.node_to_nodeid(g, trace[0])
t_idx = vft.node_to_nodeid(g, trace[-1])
sh_vf = vf_g.get_all_shortest_paths(s_idx + N, t_idx + N)
if len(sh_vf) > 0:
in_customer += 1
if first_hop_type != LinkDir.D:
in_customer_but_use_up += 1
if first_hop_type == LinkDir.D:
# 'remove' all downward path:
for neighbor in vf_g.neighbors(s_idx + N, mode=igraph.OUT):
down_edge = vf_g.get_eid(s_idx + N, neighbor, directed=True)
vf_g.es[down_edge]['LPweight_old'] = vf_g.es[down_edge][
'LPweight']
vf_g.es[down_edge]['LPweight'] = vf_sum_lp_w + 1
lp_sh = vft.get_shortest_vf_route(g,
s_idx,
t_idx,
mode='lp',
vf_g=vf_g,
vfmode=vft.CLOSENESS)
pretty_plotter.pretty_plot(g, trace, lp_sh, [])
first_new_edge = [lp_sh[0], lp_sh[1]]
first_new_hop_type = vft.edge_dir(g,
first_new_edge,
vfmode=vft.CLOSENESS)
if first_hop_type != first_new_hop_type:
out_instead_core += 1
print ''
print "Original trace: %s" % trace
print 'Original trace dir: %s' % vft.trace_to_string(g, trace)
print 'Original closeness: %s' % [
g.vs.find(x)['closeness'] for x in trace
]
print 'LP trace: %s' % [g.vs[x]['name'] for x in lp_sh]
print 'LP dir: %s' % vft.trace_to_string(g, lp_sh)
print 'LP closeness: %s' % [
g.vs.find(x)['closeness'] for x in lp_sh
]
# raw_input()
for neighbor in vf_g.neighbors(s_idx + N, mode=igraph.OUT):
down_edge = vf_g.get_eid(s_idx + N, neighbor, directed=True)
vf_g.es[down_edge]['LPweight'] = vf_g.es[down_edge][
'LPweight_old']
elif first_hop_type == LinkDir.U:
lp_sh = vft.get_shortest_vf_route(g,
s_idx,
t_idx,
mode='lp',
vf_g=vf_g,
vfmode=vft.CLOSENESS)
lp_dir = vft.trace_to_string(g, lp_sh)
# if lp_dir[1:].startswith('U') or lp_dir[1:].startswith('D'):
# continue
first_new_edge = [lp_sh[0], lp_sh[1]]
first_new_hop_type = vft.edge_dir(g,
first_new_edge,
vfmode=vft.CLOSENESS)
if first_hop_type != first_new_hop_type:
core_instead_out += 1
print ''
print "Original trace: %s" % trace
print 'Original trace dir: %s' % vft.trace_to_string(g, trace)
print 'Original closeness: %s' % [
g.vs.find(x)['closeness'] for x in trace
]
print 'LP trace: %s' % [g.vs[x]['name'] for x in lp_sh]
print 'LP dir: %s' % vft.trace_to_string(g, lp_sh)
print 'LP closeness: %s' % [
g.vs.find(x)['closeness'] for x in lp_sh
]
pretty_plotter.pretty_plot(g, trace, lp_sh, [])
# raw_input()
logger.info('Core instead down: %d' % core_instead_out)
logger.info('Down instead core: %d' % out_instead_core)
logger.info('In customer cone: %d' % in_customer)
logger.info('In customer but use UP: %d' % in_customer_but_use_up)
def main():
parser = argparse.ArgumentParser(
description=('Syntetic route generator'),
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--progressbar', action='store_true')
parser.add_argument('--verbose', '-v', action='count', default=0)
parser.add_argument('network')
parser.add_argument('meta')
parser.add_argument('all_trace_out', metavar='all-trace-out')
parser.add_argument('syntetic_out', metavar='syntetic-out')
parser.add_argument('--trace-count',
'-tc',
type=int,
dest='trace_count',
default=5000)
parser.add_argument('--random-sample',
dest='random_sample',
action='store_true')
parser.add_argument('--closeness-error',
'-ce',
type=float,
dest='closeness_error',
default=0.0)
parser.add_argument('--core-limit-percentile',
'-cl',
type=int,
dest='core_limit',
default=0)
parser.add_argument('--toggle-node-error-mode', action='store_true')
arguments = parser.parse_args()
arguments.verbose = min(len(helpers.LEVELS) - 1, arguments.verbose)
logging.getLogger('compnet').setLevel(helpers.LEVELS[arguments.verbose])
show_progress = arguments.progressbar
g = helpers.load_network(arguments.network)
g = g.simplify()
meta = helpers.load_from_json(arguments.meta)
if arguments.random_sample:
random.shuffle(meta)
meta = meta[:arguments.trace_count]
N = g.vcount()
cl = sorted([x['closeness'] for x in g.vs], reverse=True)
logger.info('Min closeness: %s' % np.min(cl))
logger.info('Max closeness: %s' % np.max(cl))
logger.info('Mean closenss: %s' % np.mean(cl))
logger.info('10%% closeness: %s' % np.percentile(cl, 10))
logger.info('90%% closeness: %s' % np.percentile(cl, 90))
logger.info('Core limit: [r]%d%%[/]' % arguments.core_limit)
change_probability = 100 * arguments.closeness_error
logger.info('Change probability: [r]%6.2f%%[/]' % change_probability)
core_limit = np.percentile(cl, arguments.core_limit)
logger.info('Core limit in closeness: [bb]%f[/]' % core_limit)
if arguments.toggle_node_error_mode:
logger.info('[r]Node error mode[/]')
msg = (
"If given node's closeness >= core_limit then the new ",
"closeness in this node is ",
#"rand(closeness_error ... old closeness)"
"OLD_CLOSENSS * +/- closeness_error%")
logger.info(''.join(msg))
logger.info('Minimum node closeness: [g]%f[/]' %
arguments.closeness_error)
for n in g.vs:
if n['closeness'] < core_limit:
continue
# sign = -1 if random.uniform(-1, 1) < 0 else 1
# n['closeness'] = n['closeness'] * (1 + sign * arguments.closeness_error)
new_closeness = random.uniform(arguments.closeness_error,
n['closeness'])
n['closeness'] = new_closeness
g_labeled = vft.label_graph_edges(g, vfmode=vft.CLOSENESS)
peer_edge_count = len([x for x in g_labeled.es if x['dir'] == LinkDir.P])
logger.info('Peer edge count: %d' % peer_edge_count)
changed_edges = []
if not arguments.toggle_node_error_mode:
msg = ("If the closeness values of the endpoints in given edge is ",
"larger than the core_limit and ",
"random(0,1) < closeness_error then change the direction ",
"for this edge")
logger.info(''.join(msg))
changed_u = changed_d = 0
changed_edges = []
changed_edgess = []
for edge in g_labeled.es:
s, t = edge.source, edge.target
s_cl = g_labeled.vs[s]['closeness']
t_cl = g_labeled.vs[t]['closeness']
if (s_cl < core_limit or t_cl < core_limit): continue
if random.uniform(0, 1) > arguments.closeness_error: continue
# if abs(s_cl - t_cl) / min(s_cl, t_cl) > 0.02: continue
new_edge_dir = LinkDir.U if random.uniform(0,
1) > 0.5 else LinkDir.D
if new_edge_dir != edge['dir']:
if edge['dir'] == LinkDir.U:
changed_u += 1
else:
changed_d += 1
edge['dir'] = new_edge_dir
changed_edges.append(edge)
changed_edgess.append((edge.source, edge.target))
# if edge['dir'] == LinkDir.U:
# changed_u += 1
# changed_edgess.append((edge.source, edge.target))
# edge['dir'] = LinkDir.D
# changed_edges.append(edge)
# elif edge['dir'] == LinkDir.D:
# changed_d += 1
# changed_edgess.append((edge.source, edge.target))
# edge['dir'] = LinkDir.U
# changed_edges.append(edge)
logger.info('E count: %d' % g_labeled.ecount())
logger.info('Changed U: %d' % changed_u)
logger.info('Changed D: %d' % changed_d)
logger.info('Changed: %d' % (changed_d + changed_u))
changed_e = [(g_labeled.vs[e.source]['name'],
g_labeled.vs[e.target]['name']) for e in changed_edges]
changed_e = changed_e + [(x[1], x[0]) for x in changed_e]
changed_e = set(changed_e)
vf_g_closeness = vft.convert_to_vf(g,
vfmode=vft.CLOSENESS,
labeled_graph=g_labeled)
# e_colors = []
# for e in vf_g_closeness.es:
# if e.source < N and e.target < N: col = 'grey'
# elif e.source < N and e.target >= N: col = 'blue'
# elif e.source >= N and e.target >= N: col = 'red'
# else: col = 'cyan'
# e_colors.append(col)
# igraph.plot(vf_g_closeness, "/tmp/closeness.pdf",
# vertex_label=vf_g_closeness.vs['name'],
# vertex_size=0.2,
# edge_color=e_colors)
pairs = [(g.vs.find(x[helpers.TRACE][0]).index,
g.vs.find(x[helpers.TRACE][-1]).index, tuple(x[helpers.TRACE]))
for x in meta]
# pairs = list(set(pairs))
# random.shuffle(pairs)
# visited = set()
# pairs2 = []
# for x in pairs:
# k = (x[0], x[1])
# if k in visited: continue
# visited.add(k)
# visited.add((k[1], k[0]))
# pairs2.append(x)
# pairs = pairs2
traces = [x[2] for x in pairs]
stretches = []
syntetic_traces = []
sh_traces = []
base_traces = []
original_traces = []
bad = 0
progress = progressbar1.DummyProgressBar(end=10, width=15)
if show_progress:
progress = progressbar1.AnimatedProgressBar(end=len(pairs), width=15)
for s, t, trace_original in pairs:
progress += 1
progress.show_progress()
trace_original_idx = [g.vs.find(x).index for x in trace_original]
logger.debug('Original trace: %s -- %s -- %s',
[g.vs[x]['name'] for x in trace_original_idx],
vft.trace_to_string(g, trace_original_idx, vft.CLOSENESS),
[g.vs[x]['closeness'] for x in trace_original_idx])
sh_routes = g.get_all_shortest_paths(s, t)
sh_len = len(sh_routes[0])
sh_routes_named = [[g.vs[y]['name'] for y in x] for x in sh_routes]
sh_trace_name = random.choice(sh_routes_named)
base_trace_name = random.choice(sh_routes_named)
candidates = vf_g_closeness.get_all_shortest_paths(s + N, t + N)
candidates = [vft.vf_route_converter(x, N) for x in candidates]
# candidates = []
if len(candidates) == 0:
candidates = vft.get_shortest_vf_route(g_labeled,
s,
t,
mode='vf',
vf_g=vf_g_closeness,
_all=True,
vfmode=vft.CLOSENESS)
if len(candidates) == 0:
s_name, t_name = g.vs[s]['name'], g.vs[t]['name']
logger.debug("!!!No syntetic route from %s to %s" %
(s_name, t_name))
continue
logger.debug('Candidates from %s to %s:' %
(g.vs[s]['name'], g.vs[t]['name']))
for c in candidates:
logger.debug('%s -- %s -- %s' %
([g.vs[x]['name'] for x in c],
vft.trace_to_string(g_labeled, c, vft.PRELABELED),
[g.vs[x]['closeness'] for x in c]))
chosen_one = random.choice(candidates)
chosen_one_name = [g.vs[x]['name'] for x in chosen_one]
# print chosen_one
# print trace_original
# pretty_plotter.pretty_plot(g, trace_original_idx,
# chosen_one, changed_edgess,
# spec_color=(0, 0, 0, 155))
hop_stretch = len(chosen_one) - sh_len
stretches.append(hop_stretch)
trace_original_e = zip(trace_original, trace_original[1:])
chosen_one_e = zip(chosen_one_name, chosen_one_name[1:])
trace_affected = any([x in changed_e for x in trace_original_e])
chosen_affected = any([x in changed_e for x in chosen_one_e])
logger.debug('Trace affected: %s' % trace_affected)
logger.debug('Chosen affected: %s' % chosen_affected)
# if hop_stretch > 2:
# logger.debug('Base: %s' % trace_to_string(g_labeled, base_trace_name))
# logger.debug('SH: %s' % trace_to_string(g_labeled, sh_trace_name))
# logger.debug('Trace: %s' % trace_to_string(g_labeled, trace_original))
# logger.debug('Syntetic: %s' % trace_to_string(g_labeled, chosen_one_name))
if trace_affected or chosen_affected or hop_stretch > 2:
# pretty_plotter.pretty_plot_all(g, traces,
# chosen_one, changed_edgess,
# spec_color=(0, 0, 0, 255))
bad += 1
syntetic_traces.append(chosen_one_name)
sh_traces.append(sh_trace_name)
base_traces.append(base_trace_name)
original_traces.append(trace_original)
logger.debug('From %s to %s chosen one %s' %
(g.vs[s]['name'], g.vs[t]['name'], chosen_one_name))
result = zip(base_traces, sh_traces, original_traces, syntetic_traces)
helpers.save_to_json(arguments.all_trace_out, result)
helpers.save_to_json(arguments.syntetic_out, syntetic_traces)
print 'Bad: %d' % bad
c = collections.Counter(stretches)
trace_count = len(syntetic_traces)
logger.info('Stretch dist:')
for k in c:
logger.info('\t%d: %5.2f%%[%d]' %
(k, 100 * c[k] / float(trace_count), c[k]))
logger.info('Valid route count: %d' % trace_count)
logger.info('Route count parameter: %d' % arguments.trace_count)
logger.info('Generated valid pair count: %d' % len(pairs))
def test_trace_to_string(self):
trace = [0, 5, 4, 6, 5, 1, 5, 6, 4, 6, 2]
trace_str = vft.trace_to_string(self.sample_graph, trace)
self.assertEqual(trace_str, 'UUDPDUPUDD')