def test_sin_cyclostationary_traffic_matrix(self):
tms = fnss.sin_cyclostationary_traffic_matrix(self.G, 10, 0.2, gamma=0.3,
log_psi=-0.3, delta=0.2,
n=24, periods=2, max_u=0.9)
self.assertEqual(48, len(tms))
self.assertAlmostEqual(0.9, max([max(fnss.link_loads(self.G, tm).values()) for tm in tms]))
self.assertLessEqual(0, min([min(fnss.link_loads(self.G, tm).values()) for tm in tms]))
def test_sin_cyclostationary_traffic_matrix_low_log_psi(self):
# Test that with very low value of log_psi and/or gamma to test
# that FNSS deals properly with division by 0 cases
log_psi = [-1500, -200]
for lp in log_psi:
if exp(lp) == 0:
self.assertRaises(ValueError,
fnss.sin_cyclostationary_traffic_matrix,
self.G, mean=10, stddev=0.2,
gamma=0.3, log_psi=lp, delta=0.2,
n=24, periods=2, max_u=0.9)
else:
try:
tms = fnss.sin_cyclostationary_traffic_matrix(
self.G, mean=10, stddev=0.2,
gamma=0.3, log_psi=lp, delta=0.2,
n=24, periods=2, max_u=0.9)
except ValueError:
pass
else:
volumes = [tm[y] for tm in tms for y in tm.flows()]
valid_volumes = [vol >= 0 and not isinf(vol) for vol in volumes]
self.assertTrue(all(valid_volumes))
weighted=False)
# add weights proportional to inverse of capacity
fnss.set_weights_inverse_capacity(topology)
# add constant link delays of 2 ms
fnss.set_delays_constant(topology, 2, delay_unit='ms')
# generate cyclostationary traffic matrix (period 7 days, 24 samples per day)
tmc = fnss.sin_cyclostationary_traffic_matrix(
topology,
mean=0.5, # average flow in TM is 0,5 Gbps
stddev=0.05, # this is the std among average flows of different OD pairs
gamma=0.8, # gamma and log_psi are parameters for fitting the std of
log_psi=-0.33, # volume fluctuations over time. Look at Nucci et al. paper
delta=0.2, # traffic variation from period max and avg as fraction of average
n=24, # number of samples per each period
periods=7, # number of periods
max_u=0.9, # max link utilization desired
origin_nodes=None, # Specify origin and destination nodes. If None,
destination_nodes=None # all nodes of the topology are both
) # origin and destination nodes of traffic
# now we generate a static traffic matrix, but this time instead of generating
# a matrix where all nodes are both origin and destinations, we pick up only
# few nodes as sources and destinations of traffic.
# Let's select 5 sources and 5 destinations
nodes = topology.nodes()
origins = random.sample(nodes, 5)
destinations = random.sample(nodes, 5)
weighted=False)
# add weights proportional to inverse of capacity
fnss.set_weights_inverse_capacity(topology)
# add constant link delays of 2 ms
fnss.set_delays_constant(topology, 2, delay_unit='ms')
# generate cyclostationary traffic matrix (period 7 days, 24 samples per day)
tmc = fnss.sin_cyclostationary_traffic_matrix(
topology,
mean=0.5, # average flow in TM is 0,5 Gbps
stddev=0.05, # this is the std among average flows of different OD pairs
gamma=0.8, # gamma and log_psi are parameters for fitting the std of
log_psi=-0.33, # volume fluctuations over time. Look at Nucci et al. paper
delta=0.2, # traffic variation from period max and avg as fraction of average
n=24, # number of samples per each period
periods=7, # number of periods
max_u=0.9, # max link utilization desired
origin_nodes=None, # Specify origin and destination nodes. If None,
destination_nodes=None # all nodes of the topology are both
) # origin and destination nodes of traffic
# now we generate a static traffic matrix, but this time instead of generating
# a matrix where all nodes are both origin and destinations, we pick up only
# few nodes as sources and destinations of traffic.
# Let's select 5 sources and 5 destinations
nodes = topology.nodes()
origins = random.sample(nodes, 5)
destinations = random.sample(nodes, 5)
