def generate_default_report(self, output_csv=False):
"""Format a default summary report based on calculated statistics for
an executed scenario.
:returns: A report (string) suitable for printing, emailing, etc.
"""
stats = self.stats
template = Template(self.scenario_template())
tmpl_vars = {
'size_data': [],
'stat_list': [
('TOTAL', stats['agg_stats'], stats['size_stats']),
('CREATE', stats['op_stats'][ssbench.CREATE_OBJECT],
stats['op_stats'][ssbench.CREATE_OBJECT]['size_stats']),
('READ', stats['op_stats'][ssbench.READ_OBJECT],
stats['op_stats'][ssbench.READ_OBJECT]['size_stats']),
('UPDATE', stats['op_stats'][ssbench.UPDATE_OBJECT],
stats['op_stats'][ssbench.UPDATE_OBJECT]['size_stats']),
('DELETE', stats['op_stats'][ssbench.DELETE_OBJECT],
stats['op_stats'][ssbench.DELETE_OBJECT]['size_stats']),
],
'agg_stats':
stats['agg_stats'],
'nth_pctile':
stats['nth_pctile'],
'start_time':
datetime.utcfromtimestamp(
stats['time_series']['start_time']).strftime(
REPORT_TIME_FORMAT),
'stop_time':
datetime.utcfromtimestamp(
stats['time_series']['stop']).strftime(REPORT_TIME_FORMAT),
'duration':
stats['time_series']['stop'] - stats['time_series']['start_time'],
'jobs_per_worker_stats':
stats['jobs_per_worker_stats'],
'weighted_c':
0.0,
'weighted_r':
0.0,
'weighted_u':
0.0,
'weighted_d':
0.0,
}
for size_data in self.scenario.sizes_by_name.values():
if size_data['size_min'] == size_data['size_max']:
size_range = '%-15s' % (self._format_bytes(
size_data['size_min']), )
else:
size_range = '%s - %s' % (
self._format_bytes(size_data['size_min']),
self._format_bytes(size_data['size_max']))
initial_files = self.scenario._scenario_data['initial_files']
initial_total = sum(initial_files.values())
pct_total = (initial_files.get(size_data['name'], 0) /
float(initial_total) * 100.0)
tmpl_vars['size_data'].append({
'crud_pcts':
' '.join(map(lambda p: '%2.0f' % p, size_data['crud_pcts'])),
'size_range':
size_range,
'size_name':
size_data['name'],
'pct_total_ops':
'%3.0f%%' % pct_total,
})
tmpl_vars['weighted_c'] += \
pct_total * size_data['crud_pcts'][0] / 100.0
tmpl_vars['weighted_r'] += \
pct_total * size_data['crud_pcts'][1] / 100.0
tmpl_vars['weighted_u'] += \
pct_total * size_data['crud_pcts'][2] / 100.0
tmpl_vars['weighted_d'] += \
pct_total * size_data['crud_pcts'][3] / 100.0
if output_csv:
csv_fields = [
'scenario_name', 'ssbench_version', 'worker_count',
'concurrency', 'start_time', 'stop_time', 'duration',
'delete_after'
]
csv_data = {
'scenario_name': self.scenario.name,
'ssbench_version': self.scenario.version,
'worker_count': tmpl_vars['agg_stats']['worker_count'],
'concurrency': self.scenario.user_count,
'start_time': tmpl_vars['start_time'],
'stop_time': tmpl_vars['stop_time'],
'duration': tmpl_vars['duration'],
'delete_after': str(self.scenario.delete_after),
}
for label, stats, sstats in tmpl_vars['stat_list']:
label_lc = label.lower()
if stats.get('req_count', 0):
self._add_csv_kv(csv_fields, csv_data,
'%s_count' % label_lc, stats['req_count'])
self._add_csv_kv(csv_fields, csv_data,
'%s_errors' % label_lc, stats['errors'])
self._add_csv_kv(csv_fields, csv_data,
'%s_retries' % label_lc, stats['retries'])
self._add_csv_kv(csv_fields, csv_data,
'%s_retry_rate' % label_lc,
'%5.2f' % stats['retry_rate'])
self._add_csv_kv(csv_fields, csv_data,
'%s_avg_req_per_s' % label_lc,
stats['avg_req_per_sec'])
self._add_stats_for(csv_fields, csv_data, label, 'all',
stats, tmpl_vars['nth_pctile'])
for size_str, per_size_stats in sstats.iteritems():
if per_size_stats:
self._add_stats_for(csv_fields, csv_data, label,
size_str, per_size_stats,
tmpl_vars['nth_pctile'])
csv_file = StringIO()
csv_writer = DictWriter(csv_file,
csv_fields,
lineterminator='\n',
quoting=csv.QUOTE_NONNUMERIC)
csv_writer.writeheader()
csv_writer.writerow(csv_data)
return csv_file.getvalue()
else:
return template.render(scenario=self.scenario, **tmpl_vars)
def parse_to_logn_and_normalize(problem_set):
columns, data = parse(problem_set)
for key in columns:
info = columns[key]
temp = list(info[:])
temp.append(len(info[1]))
columns[key] = tuple(temp)
#Calculate the attribute statistics so we can figure out the most common attribute
stats = {}
for key in data:
example = data[key]
for i in range(0, len(example)-1):
if not stats.get(i):
stats[i] = {}
if example[i]:
if stats[i].get(example[i]):
stats[i][example[i]] += 1
else:
stats[i][example[i]] = 1
#Calculate the value that should be filled in for any missing values
missing_values= {}
for i in range(0, len(example)-1):
possibles = stats[i]
if "continuous" in columns[i+1][1]:
temp = [x * y for x, y in possibles.iteritems()]
temp2 = sum(possibles.itervalues())
temp = math.fsum(temp)
missing_values[i] = temp / temp2
else:
most_common = heapq.nlargest(1, possibles.iteritems(), operator.itemgetter(1))
missing_values[i] = most_common[0][0]
#Fill in any missing values
for key in data:
example = data[key]
for i in range(0, len(example)-1):
if not example[i]:
example[i] = missing_values[i]
normalizers = {}
#Calc Normalization Params
for i in range(0, len(example)-1):
if "continuous" in columns[i+1][1]:
temp = [example[i] for example in data.itervalues()]
normalizers[i] = (statlib.stats.lmean(temp), statlib.stats.lstdev(temp))
#Create Log(N) mappings
mappings = {}
for i in range(0, len(example)-1):
if "continuous" not in columns[i+1][1]:
mappings[i] = {}
values = columns[i+1][1]
num_values = columns[i+1][2]
num_inputs = math.ceil(math.log(num_values, 2))
for j in range(0, len(values)):
mappings[i][values[j]] = [int(x) for x in tobin(j, num_inputs)]
#Convert to log(N) encoding and normalize
for key in data:
example = data[key]
new_example = []
for i in range(0, len(example)-1):
if "continuous" in columns[i+1][1]:
val = example[i]
new_val = (val - normalizers[i][0]) / normalizers[i][1]
new_example.append(new_val)
else:
new_example.extend(mappings[i][example[i]])
new_example.append(int(example[-1]))
data[key] = new_example
return(columns, data)
def generate_default_report(self, output_csv=False):
"""Format a default summary report based on calculated statistics for
an executed scenario.
:returns: A report (string) suitable for printing, emailing, etc.
"""
stats = self.stats
template = Template(self.scenario_template())
tmpl_vars = {
'size_data': [],
'stat_list': [
('TOTAL', stats['agg_stats'], stats['size_stats']),
('CREATE', stats['op_stats'][ssbench.CREATE_OBJECT],
stats['op_stats'][ssbench.CREATE_OBJECT]['size_stats']),
('READ', stats['op_stats'][ssbench.READ_OBJECT],
stats['op_stats'][ssbench.READ_OBJECT]['size_stats']),
('UPDATE', stats['op_stats'][ssbench.UPDATE_OBJECT],
stats['op_stats'][ssbench.UPDATE_OBJECT]['size_stats']),
('DELETE', stats['op_stats'][ssbench.DELETE_OBJECT],
stats['op_stats'][ssbench.DELETE_OBJECT]['size_stats']),
],
'agg_stats': stats['agg_stats'],
'nth_pctile': stats['nth_pctile'],
'start_time': datetime.utcfromtimestamp(
stats['time_series']['start_time']
).strftime(REPORT_TIME_FORMAT),
'stop_time': datetime.utcfromtimestamp(
stats['time_series']['stop']).strftime(REPORT_TIME_FORMAT),
'duration': stats['time_series']['stop']
- stats['time_series']['start_time'],
'jobs_per_worker_stats': stats['jobs_per_worker_stats'],
'weighted_c': 0.0,
'weighted_r': 0.0,
'weighted_u': 0.0,
'weighted_d': 0.0,
}
for size_data in self.scenario.sizes_by_name.values():
if size_data['size_min'] == size_data['size_max']:
size_range = '%-15s' % (
self._format_bytes(size_data['size_min']),)
else:
size_range = '%s - %s' % (
self._format_bytes(size_data['size_min']),
self._format_bytes(size_data['size_max']))
initial_files = self.scenario._scenario_data['initial_files']
initial_total = sum(initial_files.values())
pct_total = (initial_files.get(size_data['name'], 0)
/ float(initial_total) * 100.0)
tmpl_vars['size_data'].append({
'crud_pcts': ' '.join(map(lambda p: '%2.0f' % p,
size_data['crud_pcts'])),
'size_range': size_range,
'size_name': size_data['name'],
'pct_total_ops': '%3.0f%%' % pct_total,
})
tmpl_vars['weighted_c'] += \
pct_total * size_data['crud_pcts'][0] / 100.0
tmpl_vars['weighted_r'] += \
pct_total * size_data['crud_pcts'][1] / 100.0
tmpl_vars['weighted_u'] += \
pct_total * size_data['crud_pcts'][2] / 100.0
tmpl_vars['weighted_d'] += \
pct_total * size_data['crud_pcts'][3] / 100.0
if output_csv:
csv_fields = [
'scenario_name', 'ssbench_version', 'worker_count',
'concurrency', 'start_time', 'stop_time', 'duration',
'delete_after']
csv_data = {
'scenario_name': self.scenario.name,
'ssbench_version': self.scenario.version,
'worker_count': tmpl_vars['agg_stats']['worker_count'],
'concurrency': self.scenario.user_count,
'start_time': tmpl_vars['start_time'],
'stop_time': tmpl_vars['stop_time'],
'duration': tmpl_vars['duration'],
'delete_after': str(self.scenario.delete_after),
}
for label, stats, sstats in tmpl_vars['stat_list']:
label_lc = label.lower()
if stats.get('req_count', 0):
self._add_csv_kv(csv_fields, csv_data,
'%s_count' % label_lc, stats['req_count'])
self._add_csv_kv(csv_fields, csv_data,
'%s_errors' % label_lc,
stats['errors'])
self._add_csv_kv(csv_fields, csv_data,
'%s_retries' % label_lc,
stats['retries'])
self._add_csv_kv(csv_fields, csv_data,
'%s_retry_rate' % label_lc,
'%5.2f' % stats['retry_rate'])
self._add_csv_kv(csv_fields, csv_data,
'%s_avg_req_per_s' % label_lc,
stats['avg_req_per_sec'])
self._add_stats_for(csv_fields, csv_data, label, 'all',
stats, tmpl_vars['nth_pctile'])
for size_str, per_size_stats in sstats.iteritems():
if per_size_stats:
self._add_stats_for(csv_fields, csv_data, label,
size_str, per_size_stats,
tmpl_vars['nth_pctile'])
csv_file = StringIO()
csv_writer = DictWriter(csv_file, csv_fields,
lineterminator='\n',
quoting=csv.QUOTE_NONNUMERIC)
csv_writer.writeheader()
csv_writer.writerow(csv_data)
return csv_file.getvalue()
else:
return template.render(scenario=self.scenario, **tmpl_vars)