def plot_occupancy(calculated):
interpolated = {}
for n, x_val in enumerate(['length', 'size']):
nidx = pd.Float64Index(np.linspace(50, 100, 5001, endpoint=True))
for method in METHODS:
idx = calculated[method][x_val]['octets_mean']
ddd = calculated[method][x_val].copy().set_index(idx)
ddd = ddd[~ddd.index.duplicated()]
ddd = ddd.reindex(
ddd.index.union(nidx)).interpolate('slinear').reindex(nidx)
interpolated.setdefault(x_val, {})[method] = ddd
points = [99, 95, 90, 80, 75, 50]
z = pd.concat({k: pd.concat(v) for k, v in interpolated.items()})
z = z.unstack([0, 1]).swaplevel(
1, 2, axis=1).sort_index(axis=1)[['occupancy_mean', 'operations_mean']]
z = z.reindex(METHODS, axis=1, level=1)
z.loc[points][['occupancy_mean',
'operations_mean']].to_latex(f'selected.tex',
float_format='%.2f',
multicolumn_format='c')
for to in ['absolute'] + list(METHODS):
to_label = ''
fig, axes = plt.subplots(1,
2,
sharex='all',
sharey='all',
figsize=[FIGSIZE[0] * 2.132, FIGSIZE[1]])
for n, x_val in enumerate(['length', 'size']):
ax = axes[n]
for method in METHODS:
d = interpolated[x_val][method]
if to == 'absolute':
r = d.copy()
for col in r.columns:
r[col].values[:] = 1
else:
r = interpolated[x_val][to]
to_label = f' [relative to {to}]'
ax.plot(d.index,
r['occupancy_mean'] / d['occupancy_mean'],
'k' + METHODS[method],
lw=2,
label=f'{method} occupancy')
ax.plot(d.index,
r['operations_mean'] / d['operations_mean'],
'r' + METHODS[method],
lw=2 / 3,
label=f'{method} operations')
ax.set_xlabel(f'Traffic coverage [%] (decision by {x_val})')
ax.set_ylabel(f'Flow table occupancy/operations{to_label}')
ax.tick_params('y', labelleft=True)
if to == 'absolute':
ax.set_yscale('log')
ax.legend()
fig.gca().invert_xaxis()
out = f'occupancy_{to}'
save_figure(fig, out)
plt.close(fig)
def plot_probability():
fig, ax = plt.subplots(1, 1, figsize=FIGSIZE)
idx = np.geomspace(1, 1000, 512)
ax.plot(idx, 1 - (1 - 0.1)**idx, 'k-', lw=2, label='p 0.1$')
ax.plot(idx, 1 - (1 - 0.01)**idx, 'k-', lw=2, label='p = 0.1$')
ax.text(12, 0.6, '$p = 0.1$')
ax.text(150, 0.6, '$p = 0.01$')
ax.set_xlabel(f'Flow length [packets]')
ax.set_ylabel(f'Total probability of being added to flow table')
ax.set_xscale('log')
save_figure(fig, 'probability')
plt.close(fig)
def plot_traffic(calculated):
interpolated = {}
for n, x_val in enumerate(['length', 'size']):
nidx = 1 / pd.Float64Index(np.geomspace(1, 10000, 5000,
endpoint=False))
for method in METHODS:
idx = 1 / calculated[method][x_val]['occupancy_mean']
ddd = calculated[method][x_val].copy().set_index(idx)
ddd = ddd[~ddd.index.duplicated()]
ddd = ddd.reindex(
ddd.index.union(nidx)).interpolate('slinear').reindex(nidx)
interpolated.setdefault(method, {})[x_val] = ddd
for to in ['absolute'] + list(METHODS):
to_label = '%'
fig, axes = plt.subplots(1,
2,
sharex='all',
sharey='all',
figsize=[FIGSIZE[0] * 2.132, FIGSIZE[1]])
for n, x_val in enumerate(['length', 'size']):
ax = axes[n]
for method in METHODS:
d = interpolated[method][x_val]['octets_mean']
if to == 'absolute':
r = 1
else:
r = interpolated[to][x_val]['octets_mean']
to_label = f'relative to {to}'
ax.plot(d.index,
d / r,
'b' + METHODS[method],
lw=2,
label=method)
ax.set_ylabel(f'Traffic coverage [{to_label}]')
ax.set_xlabel(f'Flow table occupancy (decision by {x_val})')
ax.tick_params('y', labelleft=True)
ax.set_xscale('log')
ax.legend()
fig.gca().invert_xaxis()
out = f'traffic_{to}'
save_figure(fig, out)
save_figure(fig, out)
plt.close(fig)
def plot_all(calculated, simulated, one):
for method in calculated:
if one:
fig, axes = plt.subplots(1,
2,
figsize=[FIGSIZE[0] * 2.132, FIGSIZE[1]],
sharey='row')
txes = [ax.twinx() for ax in axes]
txes[0].get_shared_y_axes().join(*txes)
else:
fig, ax = plt.subplots(figsize=FIGSIZE)
tx = ax.twinx()
for n, x_val in enumerate(simulated[method]):
if one:
ax = axes[n]
tx = txes[n]
plot_calc_sim(ax, calculated, simulated, method, x_val, 'octets')
plot_calc_sim(tx, calculated, simulated, method, x_val, 'flows')
plot_calc_sim(tx, calculated, simulated, method, x_val, 'fraction')
ax.set_xscale('log')
tx.set_yscale('log')
ax.legend(loc=3)
tx.legend()
if method == 'sampling':
ax.invert_xaxis()
ax.set_xlabel(f'Sampling probability (sampling by {x_val})')
else:
ax.set_xlabel(f'Flow {x_val} threshold [{UNITS[x_val]}]')
if not one:
out = f'results_{method}_{x_val}'
save_figure(fig, out)
plt.close(fig)
if one:
out = f'results_{method}'
save_figure(fig, out)
plt.close(fig)