df = cmn.load_parsed_results()
def map_fn(strategy):
return 'LB-' if strategy == 'PESS' else 'TB-'
df = df[(df.isolationLevel != 'PL0') & (df.isolationLevel != 'PL1')]
df['strategy'] = df['strategy'].map(map_fn) + df['isolationLevel']
# ------ throughput
cmn.reset_colors_and_markers()
tp = df[(df.tag1 == 'ks') & (df.tag3 == 'tp')]
tp = tp.sort_values('var')
fig, ax = plt.subplots()
for key, group in tp.groupby('strategy'):
cmn.my_plot(group, ax, kind='line', x='var', y='value', label=key)
ax.set_ylim((0, 15000))
ax.margins(y=0.1)
ax.set_ylabel(cmn.TP_LABEL)
ax.set_xlabel('Number of keys')
ax.set_xscale('log')
plt.gca().invert_xaxis()
cmn.savefig_with_separate_legend('ks_tp', ax, fig)
# ------ latency
cmn.reset_colors_and_markers()
lat = df[(df.tag1 == 'ks') & (df.tag3 == 'lat')]
lat = lat.sort_values('var')
fig, ax = plt.subplots()
def get_xlabel(typee):
return 'Number of stateful operators' if 'SINGLE' in typee else 'Number of t-graphs'
# -------- Throughput
cmn.reset_colors_and_markers()
tp = df[df.tplat == 'tp'].sort_values('var')
i = 0
for typee, group in tp.groupby('type'):
fig, ax = plt.subplots()
for strategy, grp in group.groupby('strategy'):
cmn.my_plot(grp,
ax,
kind='line',
x='var',
y='value',
label=strategy)
ax.set_ylim((0, 15000))
ax.margins(y=0.1)
ax.set_xticks(range(1, 6))
ax.set_ylabel(cmn.TP_LABEL)
ax.set_xlabel(get_xlabel(typee))
i += 1
cmn.savefig_with_separate_legend('topologies_' + typee.lower() + '_tp',
ax, fig)
# -------- Latency
cmn.reset_colors_and_markers()
import pandas as pd
import matplotlib.pyplot as plt
# ------------ MAIN ------------
if __name__ == '__main__':
df = cmn.load_parsed_results()
df = df[df['tag1'] == 'scale']
df['tplat'] = df['tag3']
# -------- Throughput
cmn.reset_colors_and_markers()
tp = df[df.tplat == 'tp'].sort_values('var')
fig, ax = plt.subplots()
cmn.my_plot(tp, ax, kind='line', x='var', y='value', legend=False)
ax.set_ylim((0, 15000))
ax.margins(y=0.1)
ax.set_ylabel(cmn.TP_LABEL)
ax.set_xlabel('Graph parallelism')
cmn.savefig('scalability_tp', fig)
# -------- Latency
cmn.reset_colors_and_markers()
lat = df[df.tplat == 'lat'].sort_values('var')
fig, ax = plt.subplots()
cmn.my_plot(lat, ax, kind='line', x='var', y='value', legend=False)
accs['recovery-time-at-source']['mean']))
data = dict(replayed=x, at_state=at_state, at_source=at_source)
df = pd.DataFrame(data=data).sort_values('replayed')
cmn.reset_colors_and_markers()
fig, ax = plt.subplots()
formatter = FuncFormatter(lambda x, pos: int(x / 1000))
ax.yaxis.set_major_formatter(formatter)
ax.xaxis.set_major_formatter(formatter)
cmn.my_plot(df,
ax,
kind='line',
x='replayed',
y='at_state',
label='stateful operator',
yformatter=False)
cmn.my_plot(df,
ax,
kind='line',
x='replayed',
y='at_source',
label='open operator',
yformatter=False)
ax.set_ylim((0, 33000))
ax.margins(y=0.1)
ax.set_ylabel('Replay time [s]')
ax.set_xlabel('Number of entries in the WAL (thousands)')
pd.concat([
pd.Series(name),
group.lat.describe().rename(name),
]) for name, group in grouped)
lat.rename(columns={lat.columns[0]: 'var', 'mean': 'value'}, inplace=True)
#print tp
#print lat
#sys.exit(1)
# ------ throughput
cmn.reset_colors_and_markers()
fig, ax = plt.subplots()
#for key, group in df.groupby('percentage'):
# cmn.my_plot(group, ax, kind='line', x='percentage', y='tp', label=key)
cmn.my_plot(tp, ax, kind='line', x='var', y='value')
#ax.set_ylim((0, 80000))
ax.set_xlim((-10, 110))
ax.margins(y=0.1)
ax.set_ylabel("Average throughput [Ktxn/s]")
ax.set_xlabel('Single partition transactions percentage')
ax.set_yscale('log')
#plt.gca().invert_yaxis()
ax.get_legend().remove()
cmn.savefig('volt_tp', fig)
# ------ latency
cmn.reset_colors_and_markers()
fig, ax = plt.subplots()