def plot_outboxes(self, print_node=None, annotation=None):
thesismain.init_plot()
# plot outboxes
stats = self.csv.name.str.startswith('OutboxLength')
outbox_lengths = self.csv[self.run & self.modules & stats]
for row in outbox_lengths.itertuples():
node = row.module.split('.')[1]
outbox = row.name.split('.')[1]
if print_node is None or node == print_node:
plt.plot(row.vectime, row.vecvalue, label=outbox, linewidth=0.9, markevery=0.1)
if outbox == 'node[1]':
print(len(row.vectime))
print(row.vecvalue[-1])
plt.ylim(top=260)
axes = plt.axes()
axes.set_yticks([0, 100, 200, 250])
if annotation == 'LowHealthAttack':
plt.annotate('dropped', xy=(17, 250), xycoords='data',
xytext=(25, 200), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.ylabel('messages')
plt.xlabel('time (s)')
plt.legend(bbox_to_anchor=(-0.12, 1.1, 1.14, .102), loc='lower left', ncol=5, mode="expand")
thesismain.save_plot('plots/%s_%s_%s' % ('attack', 'outboxes', self.config_name))
def thesismain_micro_available_processing(self, annotation=None):
thesismain.init_plot()
stats = self.csv.name == 'AvailableProcessingRate'
available_processing = self.csv[self.run & self.modules & stats]
for row in available_processing.itertuples():
node = row.module.split('.')[1]
# find max's index and max value
# print(np.where(row.vecvalue == row.vecvalue.max())[0][0], row.vecvalue.max())
plt.plot(row.vectime, row.vecvalue, label=node, linewidth=0.9, markevery=0.1)
# -> 89 845.0
if annotation == 'Scenario2':
plt.ylim(bottom=-5, top=310)
plt.annotate('845 MPS', xy=(89, 302), xycoords='data',
xytext=(129, 299), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
else:
plt.ylim(bottom=-5)
plt.ylabel('throughput (MPS)')
plt.xlabel('time (s)')
thesismain.save_plot('plots/%s_%s_%s' % ('micro', 'available-processing', self.config_name))
def thesismain_micro_latency(self):
msgs_df = self.all_messages.copy()
difference_suffix = '_difference'
# calculate latency for every message on every node
for node in self.all_nodes:
msgs_df[node + difference_suffix] = msgs_df[node].subtract(msgs_df['time'])
cdfs = {}
nodes_in_sync = []
msgs_count = msgs_df.shape[0]
for node in self.all_nodes:
cdf_df = self.calculate_cdf(msgs_df, node)
cdfs[node] = cdf_df
msgs_node_missing = msgs_df[node].isnull().sum()
if msgs_node_missing < msgs_count * 0.05:
nodes_in_sync.append(node)
thesismain.init_plot()
# plot
for node in self.all_nodes:
cdf_df = cdfs[node]
plt.plot(cdf_df[node + difference_suffix], cdf_df['cdf'], label=node, linewidth=0.9, markevery=0.2)
plt.ylim(bottom=-0.02, top=1.02)
plt.xlim(left=-0.2, right=6.2)
plt.ylabel('cdf')
plt.xlabel('time (s)')
thesismain.save_plot('plots/%s_%s_%s' % ('micro', 'latency', self.config_name))
def thesismain_micro_inbox_solidification_buffer(self, annotation=None):
thesismain.init_plot((16, 4))
stats = self.csv.name == 'TotalInbox'
inbox_lengths = self.csv[self.run & self.modules & stats]
for row in inbox_lengths.itertuples():
node = row.module.split('.')[1]
plt.plot(row.vectime, row.vecvalue, label=node, linewidth=0.9, markevery=0.5)
# find a specific point in a node's graph
# if node == 'node[4]':
# print(np.where(row.vecvalue == 250)[0][-1])
# print(row.vectime[np.where(row.vecvalue == 250)[0][-1]])
if annotation == 'Scenario2':
plt.annotate('out-of-sync', xy=(54, 250), xycoords='data',
xytext=(75, 190), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.annotate('out-of-sync', xy=(120, 250), xycoords='data',
xytext=(135, 190), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.ylim(bottom=-5, top=260)
plt.ylabel('messages')
plt.xlabel('time (s)')
plt.legend(bbox_to_anchor=(-0.17, 1.1, 1.2, .102), loc='lower left', ncol=10, mode="expand")
thesismain.save_plot('plots/%s_%s_%s' % ('micro', 'inbox', self.config_name))
def plot_allowed_receiving_rate(self, print_node=None, malicious_node=None, annotation=None):
thesismain.init_plot()
# plot inboxes
stats = self.csv.name == 'AllowedReceivingRate'
allowed_receiving_rate = self.csv[self.run & self.modules & stats]
print_node_values = None
for row in allowed_receiving_rate.itertuples():
node = row.module.split('.')[1]
if print_node is None or print_node == node:
print_node_values = row.vecvalue
plt.plot(row.vectime, row.vecvalue, label='allowed', linewidth=1.7, markevery=0.2)
# plot actual receiving rate
stats = self.csv.name.str.startswith('ReceivingRate')
actual_receiving_rates = self.csv[self.run & self.modules & stats]
for row in actual_receiving_rates.itertuples():
receiver = row.module.split('.')[1]
sender = row.name.split('.')[1]
if print_node is None or print_node == receiver:
# find a specific point in a node's graph
# if sender == 'node[7]':
# print(len(row.vectime))
# print(row.vecvalue[-1])
# if malicious_node == sender:
# for count, val in enumerate(row.vecvalue):
# if len(print_node_values) > count:
# if val > print_node_values[count]:
# print('%d: %0.2f %0.2f <--' % (count, print_node_values[count], val))
# else:
# print('%d: %0.2f %0.2f' % (count, print_node_values[count], val))
plt.plot(row.vectime, row.vecvalue, label=('%s' % (sender)), linewidth=0.9)
# plt.ylim(bottom=-5, top=260)
if annotation == 'ExceedSendingRateAttack':
plt.annotate('dropped', xy=(20, 110.66), xycoords='data',
xytext=(27, 60), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
elif annotation == 'SubceedSendingRateAttack':
plt.annotate('dropped', xy=(48, 10), xycoords='data',
xytext=(60, 40), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.ylabel('rate (MPS)')
plt.xlabel('time (s)')
# plt.legend(bbox_to_anchor=(-0.12, 1.1, 1.14, .102), loc='lower left', ncol=5, mode="expand")
thesismain.save_plot('plots/%s_%s_%s' % ('attack', 'receiving_rate', self.config_name))
def thesismain_micro_throughput(self, annotation=None):
thesismain.init_plot()
msgs_df = self.all_messages.copy()
# determine how many nodes were in sync and average
total_count = None
total_in_sync_nodes = None
total_division = None
# plot nodes
for node in self.all_nodes:
label = node
count, division = np.histogram(msgs_df[node], bins=range(self.simulation_time))
plt.plot(division, np.concatenate(([0], count)), linewidth=0.9, markevery=0.1)
# find a specific point in a node's graph
# if node == 'node[6]':
# print(np.where(count == 0))
# sum up for throughput average
if total_count is not None:
total_count += count
total_in_sync_nodes += (count > 0).astype(int)
else:
total_count = count
total_in_sync_nodes = (count > 0).astype(int)
total_division = division
# determine average
total_count = total_count / total_in_sync_nodes
# plot average
plt.plot(total_division, np.concatenate(([0], total_count)), label='mean', linewidth=1.3, markevery=0.1)
# print mean value of average
print('Throughput avg mean: %.2f' % np.mean(total_count))
if annotation == 'Scenario1':
plt.annotate('out-of-sync', xy=(43, 0), xycoords='data',
xytext=(85, 25), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
elif annotation == 'Scenario2':
plt.annotate('out-of-sync', xy=(54, 0), xycoords='data',
xytext=(96, 25), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.annotate('out-of-sync', xy=(120, 0), xycoords='data',
xytext=(166, 25), textcoords='data',
arrowprops=dict(arrowstyle="->"),
horizontalalignment='right', verticalalignment='top',
)
plt.ylim(bottom=-5, top=160)
plt.ylabel('throughput (MPS)')
plt.xlabel('time (s)')
# plt.legend(bbox_to_anchor=(-0.12, 1.1, 1.14, .102), loc='lower left', ncol=5, mode="expand")
plt.legend()
thesismain.save_plot('plots/%s_%s_%s' % ('micro', 'throughput', self.config_name))