def plot_peak_ratio_cdf(rperdev1, rperdev2, agg_param, PLOTPATH):
"""
plot CDF of peak-ratio all, or peak-ratio per device
"""
fig1, ax1 = plt.subplots(1, 1)
# x-axis: peak-ratio per device, y-axis: agg (param) ratio over days
x1,y1 = getSortedCDF(rperdev1.values)
x2,y2 = getSortedCDF(rperdev2.values)
ax1.plot(x1, y1, marker='o', markevery=len(y1)//10, linestyle='--', label='treatment')
ax1.plot(x2, y2, marker='d', markevery=len(y2)//10, linestyle='--', label='control')
filename_label = agg_param.upper()
#ax1.set_xlabel(agg_param + ' peak-ratio per device')
ax1.set_xlabel('Peak-ratio per household')
#ax1.set_xscale('log')
#ax1.set_title("Distribution of peak-ratio (95%:average) per device aggregated "+agg_param+" over days")
plotname = 'peakratio-CDF-devices-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
#fig1.savefig(PLOTPATH + plotname)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_cdf_all_bytes(test_full, control_full, PLOTPATH):
fig1, ax1 = plt.subplots(1,1)
x1, y1 = getSortedCDF( test_full['octets_passed'].values )
ax1.plot(x1, y1, marker='o', color='b', markevery=len(y1)//10, linestyle='--', label='treatment')
x2, y2 = getSortedCDF( control_full['octets_passed'].values )
ax1.plot(x2, y2, marker='d', color='g', markevery=len(y2)//10, linestyle='--', label='control')
#format_axes(ax1)
#ax1.set_xscale('log')
ax1.set_xlabel("Bytes Transferred")
#ax1.set_ylabel('CDF')
#ax1.set_title('All Bytes')
plotname = 'cdf-all-bytes'
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname +".pdf", format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE " + PLOTPATH + plotname)
plt.close()
return
def plot_peak_ratio_timeseries(rperday1, rperday2, agg_param, PLOTPATH):
"""
plot timeseries of peak-ratio (mean, max, median of devices) per day
"""
fig1, ax1 = plt.subplots(1, 1)
# x-axis: DAY, y-axis: agg (param) ratio over devices
rperday1.plot(ax=ax1, marker='o', linestyle='--', label='treatment')
rperday2.plot(ax=ax1, marker='d', linestyle='--', label='control')
filename_label = agg_param.upper()
ax1.set_ylabel(agg_param+' peak-ratio')
ax1.set_yscale('log')
#ax1.set_title("Daily peak-ratio (95%:average) aggregated "+agg_param+" over devices")
plotname = 'peakratio-timeseries-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
#fig1.savefig(PLOTPATH + plotname)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_initial_timeseries(g1, g2, param, PLOTPATH):
"""
plot timeseries of data rate (mean, max, median of devices) per datetime
Input: pandas grouped by datetime; Index = datetime; Columns = throughput (over all devices)
"""
fig1, ax1 = plt.subplots(1, 1, figsize=(18,5))
# get the right timeseries
if param in ['sum', 'max', 'min', 'median', 'mean']:
ts1 = getattr(g1['throughput'], param)()
ts2 = getattr(g2['throughput'], param)()
elif param == 'perc90':
ts1 = getattr(g1['throughput'], 'quantile')(0.95)
ts2 = getattr(g2['throughput'], 'quantile')(0.95)
param = 'perc95'
else:
logger.warning("unknown parameter to plot throughput timeseries")
plt.close()
return
# plot the time series
ts1.plot(ax=ax1, marker='o', alpha=.5, linestyle='', markersize=4, label='treatment')
ts2.plot(ax=ax1, marker='d', alpha=.5, linestyle='', markersize=4, label='control')
# save with a filename containing the aggregation parameter
filename_label = param.upper()
ax1.set_ylabel('Data Rate [kbps]')
ax1.set_title(param+' Avg Data Rate')
plotname = 'timeseries-throughput-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
#fig1.savefig(PLOTPATH + plotname)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_prevalence_total_devices(test_full, control_full, PLOTPATH):
maxThresh = max( test_full['octets_passed'].max(), control_full['octets_passed'].max() )
minThresh = min( test_full['octets_passed'].min(), control_full['octets_passed'].min() )
stepThresh = (maxThresh - minThresh)/20
fig1, ax1 = plt.subplots(1,1)
ctr = 0
c = ['b', 'g']
m = ['o', 'd']
lab = ['treatment', 'control']
for df in [test_full, control_full]:
xdata = []
ydata = []
for THRESH in np.arange(minThresh, maxThresh, stepThresh):
xdata.append(THRESH)
sliced_df = df [ df['octets_passed'] >= THRESH ]
num_dev = len( sliced_df['Device_number'].unique() )
ydata.append( num_dev )
ax1.plot(xdata, ydata, color=c[ctr], marker=m[ctr], label=lab[ctr])
ctr+=1
ax1.set_xscale('linear')
ax1.set_xlabel('Threshold Bytes')
ax1.set_ylabel('Number of Devices')
ax1.set_yscale('log')
#ax1.set_title("Prevalence: total devices")
plotname = 'slice-dataset-threshold-count-ndevices'
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname + '.pdf', format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_cdf_per_device(test_full, control_full, PLOTPATH, groupBy=None, param1='max', param2='perc95'):
'''
input: test, control df ['throughput', 'date, 'day', 'month']
groupBy = None, date
param1 = max, perc95
param2 = mean, median
'''
## CDF max per device
## 95perc per device
fig1, ax1 = plt.subplots(1,1)
ctr = 0
c = ['b', 'g', 'k', 'r']
m = ['o', 'd']
lab = ['treatment', 'control']
perc95 = lambda x: x.quantile(0.95)
for df in [test_full, control_full]:
#xdata = df.groupby('Device_number')['speed'].max()
if groupBy !=None:
g = df.groupby(['Device_number', groupBy])['octets_passed']
else:
g = df.groupby(['Device_number'])['octets_passed']
if param1 == 'perc95':
xdata = getattr(g, 'quantile')(0.95)
else:
xdata = getattr(g, param1)()
if param2 == 'perc95':
xdata2 = getattr(g, 'quantile')(0.95)
elif param2 == '':
pass
else:
xdata2 = getattr(g, param2)()
x,y = getSortedCDF(xdata)
ax1.plot(x, y, color=c[ctr], marker=m[ctr], markevery=len(y)//10, label=lab[ctr])
if param2 != '':
x,y = getSortedCDF(xdata2)
ax1.plot(x, y, color=c[ctr+2], marker=m[ctr], ls='--', markevery=len(y)//10, label=lab[ctr]+'-'+param2)
ctr+=1
#format_axes(ax1)
#ax1.set_xscale('log')
ax1.set_xlabel("Bytes Transferred")
if groupBy == 'date':
ax1.set_xlabel("Bytes Transferred per Day [kbps]")
#ax1.set_ylabel('CDF')
#ax1.set_title('Max per Device')
if param2 =='' and groupBy:
plotname = 'cdf-per-device_'+groupBy+'-'+param1
elif param2 == '':
plotname = 'cdf-per-device-'+param1
elif groupBy:
plotname = 'cdf-per-device_'+groupBy+'-'+param1+'_'+param2
else:
plotname = 'cdf-per-device-'+param1+'_'+param2
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname +'.pdf', format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_primetime_ratio_per_device(r_test, r_control, param, PLOTPATH):
# CDF
fig1, ax1 = plt.subplots(1,1)
r_test_g = r_test.groupby('Device_number')['ratio']
r_control_g = r_control.groupby('Device_number')['ratio']
tit = param
if param in ['mean', 'median', 'max']:
r_t = getattr(r_test_g, param)()
r_c = getattr(r_control_g, param)()
elif param=='perc90':
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
elif param=='all1':
logger.debug("Plot: median, perc90")
r_t2 = getattr(r_test_g, 'median')()
r_c2 = getattr(r_control_g, 'median')()
x,y = getSortedCDF(r_t2)
ax1.plot(x, y, marker='o', color='k', linestyle='--', label='treatment-median', markevery=len(y)//10)
x,y = getSortedCDF(r_c2)
ax1.plot(x, y, marker='d', color='r', linestyle='--', label='control-median', markevery=len(y)//10)
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
tit = 'median, perc95'
elif param=='all2':
logger.debug("Plot: mean, max")
r_t2 = getattr(r_test_g, 'mean')()
r_c2 = getattr(r_control_g, 'mean')()
x,y = getSortedCDF(r_t2)
ax1.plot(x, y, marker='o', color='k', linestyle='--', label='treatment-mean', markevery=len(y)//10)
x,y = getSortedCDF(r_c2)
ax1.plot(x, y, marker='d', color='r', linestyle='--', label='control-mean', markevery=len(y)//10)
r_t = getattr(r_test_g, 'max')()
r_c = getattr(r_control_g, 'max')()
tit = 'mean, max'
else:
logger.debug("Plot all: median, perc90")
r_t2 = getattr(r_test_g, 'median')()
r_c2 = getattr(r_control_g, 'median')()
x,y = getSortedCDF(r_t2)
ax1.plot(x, y, marker='o', color='k', linestyle='--', label='treatment-median', markevery=len(y)//10)
x,y = getSortedCDF(r_c2)
ax1.plot(x, y, marker='d', color='r', linestyle='--', label='control-median', markevery=len(y)//10)
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
x,y = getSortedCDF(r_t)
ax1.plot(x, y, marker='o', color='b', label='treatment', markevery=len(y)//10)
x,y = getSortedCDF(r_c)
ax1.plot(x, y, marker='d', color='g', label='control', markevery=len(y)//10)
ax1.set_xscale('log')
ax1.set_xlabel("Prime-time Ratio")
#ax1.set_ylabel('CDF')
#ax1.set_title('Prime-time Ratio per Device - '+tit)
filename_label=param.upper()
plotname = 'prime-time-ratio-per-device-cdf-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_primetime_ratio_by_date(r_test, r_control, param, PLOTPATH):
# Timeseries
fig1, ax1 = plt.subplots(1, 1, figsize=(13,8))
tit = param
lab = param
r_test_g = r_test.groupby('date')['ratio']
r_control_g = r_control.groupby('date')['ratio']
if param in ['mean', 'median', 'max']:
r_t = getattr(r_test_g, param)()
r_c = getattr(r_control_g, param)()
elif param=='perc90':
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
elif param=='all1':
logger.debug("Plot: median, perc90")
r_t2 = getattr(r_test_g, 'median')()
r_c2 = getattr(r_control_g, 'median')()
r_t2.plot(ax=ax1, marker='o', color='k', linestyle='--', label='treatment-median')
r_c2.plot(ax=ax1, marker='d', color='r', linestyle='--', label='control-median')
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
lab = 'perc95'
tit = 'median, perc95'
elif param=='all2':
logger.debug("Plot: mean, max")
r_t2 = getattr(r_test_g, 'mean')()
r_c2 = getattr(r_control_g, 'mean')()
r_t2.plot(ax=ax1, marker='o', color='k', linestyle='--', label='treatment-mean')
r_c2.plot(ax=ax1, marker='d', color='r', linestyle='--', label='control-mean')
r_t = getattr(r_test_g, 'max')()
r_c = getattr(r_control_g, 'max')()
lab = 'max'
tit = 'mean, max'
else:
logger.debug("Plot all: median, perc90")
r_t2 = getattr(r_test_g, 'median')()
r_c2 = getattr(r_control_g, 'median')()
r_t2.plot(ax=ax1, marker='o', color='k', linestyle='--', label='treatment')
r_c2.plot(ax=ax1, marker='d', color='r', linestyle='--', label='control')
r_t = getattr(r_test_g, 'quantile')(0.95)
r_c = getattr(r_control_g, 'quantile')(0.95)
r_t.plot(ax=ax1, color='b', marker='o', label='treatment')
r_c.plot(ax=ax1, color='g', marker='d', label='control')
#ax1.set_ylabel('Prime-time ratio (log)')
#ax1.set_yscale('log')
ax1.set_title("Prime-time Ratio every Date - "+tit)
filename_label=param.upper()
plotname = 'prime-time-ratio-by-date-timeseries-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return
def plot_throughput_per_day(g1, g2, param_device, param_time, PLOTPATH):
"""
Same as above, just convert octets to throughput
Input: pandas groupedby; Index=['time', 'day']; Columns=[sum, perc90, max, min, median, ...]
param_device: denotes the agg of bytes in a particular datetime over devices for the timeseries
param_time: denotes the agg over [week, time] group: median, mean, perc90, max
"""
fig1, ax1 = plt.subplots(1, 1, figsize=(18,8))
tit=param_time
if param_time in ['mean', 'max', 'min', 'median']:
ts1 = getattr(g1[param_device], param_time) * CONVERT_OCTETS
ts2 = getattr(g2[param_device], param_time) * CONVERT_OCTETS
ts1.plot(ax=ax1, color='b', linestyle='-.', linewidth=3, label='treatment-'+param_time)
ts2.plot(ax=ax1, color='g', linestyle='-.', linewidth=3, label='control-'+param_time)
elif param_time == 'perc90':
param_time = 'perc95'
ts1 = g1[param_device].quantile(0.95)* CONVERT_OCTETS
ts2 = g2[param_device],quantile(0.95)* CONVERT_OCTETS
ts1.plot(ax=ax1, color='b', linestyle='-.', linewidth=3, label='treatment-'+param_time)
ts2.plot(ax=ax1, color='g', linestyle='-.', linewidth=3, label='control-'+param_time)
elif param_time == 'all1':
logger.debug("plot perc90, median")
tit = 'perc95, median'
(g1[param_device].quantile(0.95)* CONVERT_OCTETS).plot(ax=ax1, color='b', linestyle='-', label='treatment-perc95')
(g2[param_device].quantile(0.95)* CONVERT_OCTETS).plot(ax=ax1, color='g', linestyle='-', label='control-perc95')
(g1[param_device].median()* CONVERT_OCTETS).plot(ax=ax1, color='k', linestyle='--', linewidth=2, label='treatment-median')
(g2[param_device].median()* CONVERT_OCTETS).plot(ax=ax1, color='r', linestyle='--', linewidth=2, label='control-median')
elif param_time == 'all2':
logger.debug("plot max, mean")
tit = 'max, mean'
(g1[param_device].max() * CONVERT_OCTETS).plot(ax=ax1, color='b', linestyle='-', label='treatment-max')
(g2[param_device].max() * CONVERT_OCTETS).plot(ax=ax1, color='g', linestyle='-', label='control-max')
(g1[param_device].mean()* CONVERT_OCTETS).plot(ax=ax1, color='k', linestyle='--', linewidth=2, label='treatment-mean')
(g2[param_device].mean()* CONVERT_OCTETS).plot(ax=ax1, color='r', linestyle='--', linewidth=2, label='control-mean')
else:
logger.debug("no param_time selected so plot 90-%ile, median, max, mean over time")
(g1[param_device].max() * CONVERT_OCTETS).plot(ax=ax1, color='b', linestyle='-.', linewidth=3, label='treatment')
(g2[param_device].max() * CONVERT_OCTETS).plot(ax=ax1, color='g', linestyle='-.', linewidth=3, label='control')
(g1[param_device].quantile(0.95)* CONVERT_OCTETS).plot(ax=ax1, color='k', linestyle='-', label='treatment')
(g2[param_device].quantile(0.95)* CONVERT_OCTETS).plot(ax=ax1, color='r', linestyle='-', label='control')
(g1[param_device].mean()* CONVERT_OCTETS).plot(ax=ax1, color='b', linestyle=':', linewidth=3, label='treatment')
(g2[param_device].mean()* CONVERT_OCTETS).plot(ax=ax1, color='g', linestyle=':', linewidth=3, label='control')
(g1[param_device].median()* CONVERT_OCTETS).plot(ax=ax1, color='k', linestyle='--', linewidth=2, label='treatment')
(g2[param_device].median()* CONVERT_OCTETS).plot(ax=ax1, color='r', linestyle='--', linewidth=2, label='control')
ax1.set_ylabel(param_time + '$_{time}$ '+param_device+'$_{device}$ Data Rate [kbps]')
ax1.set_title(tit + " Data Rate in a 15 min slot")
filename_label = param_time.upper()
plotname = 'describe-total-throughput-per-day-'+filename_label
ax1.grid(1)
ax1.legend(loc='best')
fig1.tight_layout()
if LATEXIFY:
format_axes(ax1)
fig1.savefig(PLOTPATH + plotname+'.pdf', format='PDF')
#fig1.savefig(PLOTPATH + plotname)
logger.info("CREATE FILE "+PLOTPATH + plotname)
plt.close()
return