1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
|
#!/usr/bin/env python2.7
import collections
import hashlib
import itertools
import json
import math
import tabulate
import time
SELF_TIME = object()
TIME_FROM_SCOPE_START = object()
TIME_TO_SCOPE_END = object()
TIME_FROM_STACK_START = object()
TIME_TO_STACK_END = object()
class LineItem(object):
def __init__(self, line, indent):
self.tag = line['tag']
self.indent = indent
self.start_time = line['t']
self.end_time = None
self.important = line['imp']
self.times = {}
class ScopeBuilder(object):
def __init__(self, call_stack_builder, line):
self.call_stack_builder = call_stack_builder
self.indent = len(call_stack_builder.stk)
self.top_line = LineItem(line, self.indent)
call_stack_builder.lines.append(self.top_line)
self.first_child_pos = len(call_stack_builder.lines)
def mark(self, line):
line_item = LineItem(line, self.indent + 1)
line_item.end_time = line_item.start_time
self.call_stack_builder.lines.append(line_item)
def finish(self, line):
assert line['tag'] == self.top_line.tag
final_time_stamp = line['t']
assert self.top_line.end_time is None
self.top_line.end_time = final_time_stamp
assert SELF_TIME not in self.top_line.times
self.top_line.times[SELF_TIME] = final_time_stamp - self.top_line.start_time
for line in self.call_stack_builder.lines[self.first_child_pos:]:
if TIME_FROM_SCOPE_START not in line.times:
line.times[TIME_FROM_SCOPE_START] = line.start_time - self.top_line.start_time
line.times[TIME_TO_SCOPE_END] = final_time_stamp - line.end_time
class CallStackBuilder(object):
def __init__(self):
self.stk = []
self.signature = hashlib.md5()
self.lines = []
def finish(self):
start_time = self.lines[0].start_time
end_time = self.lines[0].end_time
self.signature = self.signature.hexdigest()
for line in self.lines:
line.times[TIME_FROM_STACK_START] = line.start_time - start_time
line.times[TIME_TO_STACK_END] = end_time - line.end_time
def add(self, line):
line_type = line['type']
self.signature.update(line_type)
self.signature.update(line['tag'])
if line_type == '{':
self.stk.append(ScopeBuilder(self, line))
return False
elif line_type == '}':
self.stk.pop().finish(line)
if not self.stk:
self.finish()
return True
return False
elif line_type == '.' or line_type == '!':
self.stk[-1].mark(line)
return False
else:
raise Exception('Unknown line type: \'%s\'' % line_type)
class CallStack(object):
def __init__(self, initial_call_stack_builder):
self.count = 1
self.signature = initial_call_stack_builder.signature
self.lines = initial_call_stack_builder.lines
for line in self.lines:
for key, val in line.times.items():
line.times[key] = [val]
def add(self, call_stack_builder):
assert self.signature == call_stack_builder.signature
self.count += 1
assert len(self.lines) == len(call_stack_builder.lines)
for lsum, line in itertools.izip(self.lines, call_stack_builder.lines):
assert lsum.tag == line.tag
assert lsum.times.keys() == line.times.keys()
for k, lst in lsum.times.iteritems():
lst.append(line.times[k])
def finish(self):
for line in self.lines:
for lst in line.times.itervalues():
lst.sort()
builder = collections.defaultdict(CallStackBuilder)
call_stacks = collections.defaultdict(CallStack)
print 'Loading...'
lines = 0
start = time.time()
with open('latency_trace.txt') as f:
for line in f:
lines += 1
inf = json.loads(line)
thd = inf['thd']
cs = builder[thd]
if cs.add(inf):
if cs.signature in call_stacks:
call_stacks[cs.signature].add(cs)
else:
call_stacks[cs.signature] = CallStack(cs)
del builder[thd]
time_taken = time.time() - start
print 'Read %d lines in %f seconds (%f lines/sec)' % (lines, time_taken, lines / time_taken)
print 'Analyzing...'
call_stacks = sorted(call_stacks.values(), key=lambda cs: cs.count, reverse=True)
for cs in call_stacks:
cs.finish()
print 'Writing report...'
def percentile(N, percent, key=lambda x:x):
"""
Find the percentile of a list of values.
@parameter N - is a list of values. Note N MUST BE already sorted.
@parameter percent - a float value from 0.0 to 1.0.
@parameter key - optional key function to compute value from each element of N.
@return - the percentile of the values
"""
if not N:
return None
k = (len(N)-1) * percent
f = math.floor(k)
c = math.ceil(k)
if f == c:
return key(N[int(k)])
d0 = key(N[int(f)]) * (c-k)
d1 = key(N[int(c)]) * (k-f)
return d0+d1
def tidy_tag(tag):
if tag[0:10] == 'GRPC_PTAG_':
return tag[10:]
return tag
def time_string(values):
num_values = len(values)
return '%.1f/%.1f/%.1f' % (
1e6 * percentile(values, 0.5),
1e6 * percentile(values, 0.9),
1e6 * percentile(values, 0.99))
def time_format(idx):
def ent(line, idx=idx):
if idx in line.times:
return time_string(line.times[idx])
return ''
return ent
FORMAT = [
('TAG', lambda line: '..'*line.indent + tidy_tag(line.tag)),
('FROM_STACK_START', time_format(TIME_FROM_STACK_START)),
('SELF', time_format(SELF_TIME)),
('TO_STACK_END', time_format(TIME_TO_STACK_END)),
('FROM_SCOPE_START', time_format(TIME_FROM_SCOPE_START)),
('SELF', time_format(SELF_TIME)),
('TO_SCOPE_END', time_format(TIME_TO_SCOPE_END)),
]
for cs in call_stacks:
print cs.count
header, _ = zip(*FORMAT)
table = []
for line in cs.lines:
fields = []
for _, fn in FORMAT:
fields.append(fn(line))
table.append(fields)
print tabulate.tabulate(table, header, tablefmt="simple")
|
