def plot_runtime(algebras, data, output_name):
# plot wall clock time
runtime = [] # query runtime
std = [] # standard deviation
for row in data:
if row[0] == query:
runtime.append(to_float(row[2]))
std.append(to_float(row[3]))
ind = np.arange(len(runtime)) # the x locations for the groups
fig, ax = plt.subplots()
rects = ax.bar(ind+0.1, runtime, width=bar_width, color=colors, yerr=std)
ax.set_ylabel('Time (sec)')
ax.set_xticks(ind+xstick_offset)
ax.set_xticklabels(algebras)
if query == "fb_q5":
ax.set_ylim((0, 14000))
# label the bars
for rect in rects:
height = rect.get_height()
if int(height) == 0:
ax.text(
rect.get_x()+rect.get_width()/2.,
1.05*height, "FAIL",
ha='center', va='bottom', color='red')
else:
ax.text(
rect.get_x()+rect.get_width()/2.,
1.05*height, to_label(height),
ha='center', va='bottom')
print "outputing {}".format(output_name)
plt.savefig(output_name, format='pdf', dpi=dpi)
def plot_runtime(algebras, data, output_name):
# plot wall clock time
runtime = [] # query runtime
std = [] # standard deviation
for row in data:
if row[0] == query:
runtime.append(to_float(row[2]))
std.append(to_float(row[3]))
ind = np.arange(len(runtime)) # the x locations for the groups
fig, ax = plt.subplots()
rects = ax.bar(ind + 0.1, runtime, width=bar_width, color=colors, yerr=std)
ax.set_ylabel('Time (sec)')
ax.set_xticks(ind + xstick_offset)
ax.set_xticklabels(algebras)
if query == "fb_q5":
ax.set_ylim((0, 14000))
# label the bars
for rect in rects:
height = rect.get_height()
if int(height) == 0:
ax.text(rect.get_x() + rect.get_width() / 2.,
1.05 * height,
"FAIL",
ha='center',
va='bottom',
color='red')
else:
ax.text(rect.get_x() + rect.get_width() / 2.,
1.05 * height,
to_label(height),
ha='center',
va='bottom')
print "outputing {}".format(output_name)
plt.savefig(output_name, format='pdf', dpi=dpi)
def max_workload_plot(fname, ofname):
"""
Plot the maximum workload per server using different HC configs
"""
# read data from csv
with open("csvs/{}".format(fname), "rU") as f:
csvreader = csv.reader(f)
data = [list(row) for row in csvreader]
N = 4
ind = np.arange(N) # x locations
bar_width = 0.27 # the width of the bars
# prepare data
raco_hc = []
random_hc = []
lp_round = []
for row in data[1:]:
opt = to_float(row[3])
raco_hc.append(to_float(row[1])/opt)
random_hc.append(to_float(row[2])/opt)
lp_round.append(to_float(row[4])/opt)
# set bars
fig, ax = plt.subplots()
data_groups = [raco_hc, lp_round, random_hc]
bars = []
for i, group in enumerate(data_groups):
rect = ax.bar(
ind+0.1+i*bar_width, group, width=bar_width, color=colors[i])
bars.append(rect)
# set labels
ax.set_ylabel("Workload to optimal ratio")
ax.set_xticks(ind+bar_width*2)
ax.set_xticklabels(("Q1", "Q2", "Q3", "Q4"))
ax.set_ylim((0, 7))
# set bar labels
for bar in bars:
for rect in bar:
height = rect.get_height()
ax.text(
rect.get_x()+rect.get_width()/2.,
1.05*height, '%.2f' % height,
ha='center', va='bottom', size=14)
ax.legend(
(bars[0][0], bars[1][0], bars[2][0]),
('Our Alg.', 'Round Down', 'Random(4096 cells)'),
prop={'size': 15})
output_path = "{}/{}".format(path, ofname)
print "outputing {}".format(output_path)
plt.savefig(output_path, format='pdf', dpi=dpi)
def attrorder_plot():
datafile = "csvs/SIGMOD Experiment - attr_order_sum.csv"
with open(datafile, "rU") as f:
reader = csv.reader(f)
data = [list(row) for row in reader]
plt.figure()
data1 = [(to_float(row[1]), to_float(row[2])) for row in data[1:]
if row[3] == "Q3"]
cost1, time1 = zip(*data1)
data2 = [(to_float(row[1]), to_float(row[2])) for row in data[1:]
if row[3] == "Q4"]
cost2, time2 = zip(*data2)
data3 = [(to_float(row[1]), to_float(row[2])) for row in data[1:]
if row[3] == "Q7"]
cost3, time3 = zip(*data3)
data4 = [(to_float(row[1]), to_float(row[2])) for row in data[1:]
if row[3] == "Q8"]
cost4, time4 = zip(*data4)
fig, ax = plt.subplots()
ax.scatter(cost1, time1, color=colors[0], s=120, label="Q3", marker="x")
ax.scatter(cost2, time2, color=colors[1], s=120, label="Q4", marker="x")
ax.scatter(cost3, time3, color=colors[2], s=120, label="Q7", marker="x")
ax.scatter(cost4, time4, color=colors[3], s=120, label="Q8", marker="x")
ax.set_xscale('log')
ax.set_yscale('log')
plt.xlabel('Estimated cost')
plt.ylabel('Actual running time (Sec)')
ax.legend(prop={'size': 15})
plt.axis([0, 10e22, -30, 1200])
oname = "{}/attr_order_scatter.pdf".format(path)
print "output to {}".format(oname)
plt.savefig(oname, format='pdf')
def plot_shuffle(algebras, data, output_name):
shuffle_size = []
for row in data:
if row[0] == query:
shuffle_size.append(to_float(row[10]))
ind = np.arange(len(shuffle_size)) # the x locations for the groups
fig, ax = plt.subplots()
rects = ax.bar(ind+0.1, shuffle_size, width=bar_width, color=colors)
ax.set_ylabel('Tuples shuffled (million)')
# ax.set_xlabel('Physical Algebra')
ax.set_xticks(ind+xstick_offset)
ax.set_xticklabels(algebras)
if query == "fb_q5":
ax.set_ylim((0, 17000))
# label the bar
for rect in rects:
height = rect.get_height()
if int(height) == 0:
ax.text(
rect.get_x()+rect.get_width()/2.,
1.05*height, "FAIL",
ha='center', va='bottom', color='red')
else:
ax.text(
rect.get_x()+rect.get_width()/2.,
1.05*height, to_label(height),
ha='center', va='bottom')
print "outputing {}".format(output_name)
plt.savefig(output_name, format='pdf', dpi=dpi)
def plot_shuffle(algebras, data, output_name):
shuffle_size = []
for row in data:
if row[0] == query:
shuffle_size.append(to_float(row[10]))
ind = np.arange(len(shuffle_size)) # the x locations for the groups
fig, ax = plt.subplots()
rects = ax.bar(ind + 0.1, shuffle_size, width=bar_width, color=colors)
ax.set_ylabel('Tuples shuffled (million)')
# ax.set_xlabel('Physical Algebra')
ax.set_xticks(ind + xstick_offset)
ax.set_xticklabels(algebras)
if query == "fb_q5":
ax.set_ylim((0, 17000))
# label the bar
for rect in rects:
height = rect.get_height()
if int(height) == 0:
ax.text(rect.get_x() + rect.get_width() / 2.,
1.05 * height,
"FAIL",
ha='center',
va='bottom',
color='red')
else:
ax.text(rect.get_x() + rect.get_width() / 2.,
1.05 * height,
to_label(height),
ha='center',
va='bottom')
print "outputing {}".format(output_name)
plt.savefig(output_name, format='pdf', dpi=dpi)