-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtraceStats.py
executable file
·499 lines (416 loc) · 17.1 KB
/
traceStats.py
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
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
#! /usr/bin/env python
# Copyright 2007,2008 The University of New South Wales
# Author: Joshua Root <[email protected]>
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the Australian Public Licence B. See the file
# OZPLB.txt for the licence terms.
"""
Calculate statistics for a trace, and optionally, use them to create an
fio job file that approximates the workload seen in the trace.
"""
# use psyco JIT if available (only on IA-32...)
try:
import psyco
psyco.full()
except ImportError:
pass
from getopt import gnu_getopt
import scipy.stats.distributions
import math
import re
import sys
sectorSize = 512
bigmem = False #setting True keeps trace data in memory rather than two-pass streaming
do_fio = False # output an fio job file that approximates the commands' access patterns
def parseArgs():
"""Handle command-line options."""
global infilename, bigmem, do_fio
optlist, args = gnu_getopt(sys.argv[1:], "i:mf")
for opt,val in optlist:
if opt == "-i":
infilename = val
elif opt == "-m":
bigmem = True
elif opt == "-f":
do_fio = True
else:
print "Unknown option: "+opt
sys.exit(2)
if infilename is None:
print "Usage: traceStats.py -i infile [-m] [-f]"
sys.exit(2)
class TSVals(object):
"""Values for a particular process"""
def __init__(self):
if bigmem:
self.ops = []
self.totalOps = 0
self.lastTime = 0.0
self.lastCompletion = 0.0
self.lastSector = 0
self.depth = 0
self.maxDepth = 0
self.cweight = 1 # how many requests will actually be completed
# by the next C event
self.reads = 0
self.writes = 0
self.lastWasRead = True
self.dirSwaps = 0
self.dirSwapTime = 0.0
self.lastDirSwap = 0.0
self.sync = 0
self.async = 0
self.maxRead = -sys.maxint-1
self.maxWrite = -sys.maxint-1
self.minRead = sys.maxint
self.minWrite = sys.maxint
self.maxSeek = -sys.maxint-1
self.minSeek = sys.maxint
self.seekSum = 0
self.sequentialCount = 0
self.minSize = sys.maxint
self.maxSize = 0
self.sizeSum = 0
self.minDelay = float(sys.maxint)
self.maxDelay = 0.0
self.delaySum = 0.0
self.meanThink = 0.0
self.szVar = 0
self.skVar = 0
self.dlVar = 0
self.nbins = 1000 # yeah, that's "lots", right?
def initBins(self):
self.szBins = {}
self.szBinWidth = float(self.maxSize-self.minSize)/(self.nbins-1)
if self.szBinWidth == 0.0:
self.szBinWidth = float(self.maxSize)
self.skBins = {}
self.skBinWidth = float(self.maxSeek-self.minSeek)/(self.nbins-1)
if self.skBinWidth == 0.0:
self.skBinWidth = float(self.maxSeek)
self.dlBins = {}
self.dlBinWidth = (self.maxDelay-self.minDelay)/(self.nbins-1)
if self.dlBinWidth == 0.0:
self.dlBinWidth = float(self.maxDelay)
for bin in range(self.nbins):
self.szBins[bin] = 0
self.skBins[bin] = 0
self.dlBins[bin] = 0
def calcChi2(self):
szdist = {}
szdist["norm"] = lambda x:scipy.stats.distributions.norm.cdf( \
x,loc=self.meanSize,scale=self.szDev)
if self.szVar == 0.0:
self.szVar = 0.001 #HACK
sztheta = self.szVar / self.meanSize
szk = self.meanSize / sztheta
szdist["gamma"] = lambda x:scipy.stats.distributions.gamma.cdf( \
x, szk, scale=sztheta)
szdist["poisson"] = lambda x:scipy.stats.distributions.poisson.cdf( \
int(round(x)), self.meanSize)
szdist["uniform"] = lambda x:scipy.stats.distributions.uniform.cdf( \
x, loc=self.minSize, scale=(self.maxSize-self.minSize))
skdist = {}
skdist["norm"] = lambda x:scipy.stats.distributions.norm.cdf( \
x,loc=self.meanSeek,scale=self.skDev)
if self.skVar == 0.0:
self.skVar = 0.001 #HACK
sktheta = self.skVar / self.meanSeek
skk = self.meanSeek / sktheta
skdist["gamma"] = lambda x:scipy.stats.distributions.gamma.cdf( \
x, skk, scale=sktheta)
skdist["poisson"] = lambda x:scipy.stats.distributions.poisson.cdf( \
int(round(x)), self.meanSeek)
skdist["uniform"] = lambda x:scipy.stats.distributions.uniform.cdf( \
x, loc=self.minSeek, scale=(self.maxSeek-self.minSeek))
dldist = {}
dldist["norm"] = lambda x:scipy.stats.distributions.norm.cdf( \
x,loc=self.meanDelay,scale=self.dlDev)
if self.dlVar == 0.0:
self.dlVar = 0.001 #HACK
dltheta = self.dlVar / self.meanDelay
dlk = self.meanDelay / dltheta
dldist["gamma"] = lambda x:scipy.stats.distributions.gamma.cdf( \
x, dlk, scale=dltheta)
dldist["poisson"] = lambda x:scipy.stats.distributions.poisson.cdf( \
int(round(x)), self.meanDelay)
dldist["uniform"] = lambda x:scipy.stats.distributions.uniform.cdf( \
x, loc=self.minDelay, scale=(self.maxDelay-self.minDelay))
szchi2 = {}
skchi2 = {}
dlchi2 = {}
for dist in ["norm","gamma","poisson","uniform"]:
szchi2[dist] = 0.0
skchi2[dist] = 0.0
dlchi2[dist] = 0.0
for bin in range(self.nbins):
# e = expected
esize = float(szdist[dist]((bin*self.szBinWidth + self.minSize)+ \
(self.szBinWidth/2))) - float(szdist[dist](( \
bin*self.szBinWidth + self.minSize)-(self.szBinWidth/2)))
if esize > 0.0:
szchi2[dist] += (self.szBins[bin] - esize)**2 / esize
eseek = float(skdist[dist]((bin*self.skBinWidth + self.minSeek)+
(self.skBinWidth/2))) - float(skdist[dist]((bin*self.skBinWidth \
+ self.minSeek)-(self.skBinWidth/2)))
if eseek > 0.0:
skchi2[dist] += (self.skBins[bin] - eseek)**2 / eseek
edelay = float(dldist[dist]((bin*self.dlBinWidth + self.minDelay)+ \
(self.dlBinWidth/2))) - float(dldist[dist]((bin*self.dlBinWidth \
+ self.minDelay)-(self.dlBinWidth/2)))
if edelay > 0.0:
dlchi2[dist] += (self.dlBins[bin] - edelay)**2 / edelay
print "size chi2("+dist+") = "+str(szchi2[dist])
print "seek chi2("+dist+") = "+str(skchi2[dist])
print "delay chi2("+dist+") = "+str(dlchi2[dist])
def handleCompletion(vals, name, thisTime):
if name not in vals:
vals[name] = TSVals()
val = vals[name]
overall = vals["__overall"]
for v in [val, overall]:
v.lastCompletion = thisTime
v.depth -= v.cweight
v.cweight = 1
if v.depth < 0:
v.depth = 0
def handleMerge(vals, name):
if name not in vals:
vals[name] = TSVals()
val = vals[name]
overall = vals["__overall"]
for v in [val, overall]:
v.cweight += 1
def updateRun1(vals, rw, sync, size, sector, thisTime, name):
if name not in vals:
vals[name] = TSVals()
val = vals[name]
overall = vals["__overall"]
for v in [val, overall]:
v.totalOps += 1
v.depth += 1
if v.depth > v.maxDepth:
v.maxDepth = v.depth
v.meanThink += thisTime - v.lastCompletion
if rw:
v.writes += 1
if v.maxWrite < size:
v.maxWrite = size
if v.minWrite > size:
v.minWrite = size
if v.lastWasRead:
v.dirSwaps += 1
v.lastWasRead = False
v.dirSwapTime += thisTime - v.lastDirSwap
v.lastDirSwap = thisTime
else:
v.reads += 1
if v.maxRead < size:
v.maxRead = size
if v.minRead > size:
v.minRead = size
if not v.lastWasRead:
v.dirSwaps += 1
v.lastWasRead = True
v.dirSwapTime += thisTime - v.lastDirSwap
v.lastDirSwap = thisTime
if sync:
v.sync += 1
else:
v.async += 1
if v.maxSize < size:
v.maxSize = size
if v.minSize > size:
v.minSize = size
v.sizeSum += size
seek = sector - v.lastSector
if v.maxSeek < seek:
v.maxSeek = seek
if v.minSeek > seek:
v.minSeek = seek
v.lastSector = sector + (size/sectorSize)
v.seekSum += seek
if seek == 0:
v.sequentialCount += 1
delay = thisTime - v.lastTime
if v.maxDelay < delay:
v.maxDelay = delay
if v.minDelay > delay:
v.minDelay = delay
v.lastTime = thisTime
v.delaySum += delay
if bigmem:
v.ops.append((rw,size,seek,delay))
def updateRun2(name, vals, size, sector, thisTime):
val = vals[name]
overall = vals["__overall"]
for v in [val, overall]:
if v.totalOps < 1:
continue
v.szVar += (size - v.meanSize)**2
#print name
v.szBins[int((size-v.minSize)/v.szBinWidth)] += 1
seek = sector - v.lastSector
v.lastSector = sector + (size/sectorSize)
v.skVar += (seek - v.meanSeek)**2
v.skBins[int((seek-v.minSeek)/v.skBinWidth)] += 1
delay = thisTime - v.lastTime
v.lastTime = thisTime
v.dlVar += (delay - v.meanDelay)**2
v.dlBins[int((delay-v.minDelay)/v.dlBinWidth)] += 1
def calcMeans(vals):
for name in vals:
v = vals[name]
if v.totalOps != 0:
pw = float(v.writes) / v.totalOps
pr = float(v.reads) / v.totalOps
v.meanSize = float(v.sizeSum) / v.totalOps
v.meanSeek = float(v.seekSum*sectorSize) / v.totalOps
v.meanDelay = v.delaySum / v.totalOps
v.meanThink /= v.totalOps
if not do_fio:
print name+":"
print str(v.reads)+" reads, "+str(v.writes)+" writes"
print "r/w ratio: "+str(pr)+"/"+str(pw)
print "size min/mean/max (B) = "+str(v.minSize)+"/"+str(v.meanSize) \
+"/"+str(v.maxSize)
print "seek min/mean/max (B) = "+str(v.minSeek)+"/"+str(v.meanSeek) \
+"/"+str(v.maxSeek)
print "delay min/mean/max (sec) = "+str(v.minDelay)+"/"+str(v.meanDelay) \
+"/"+str(v.maxDelay)
print ""
if v.dirSwaps != 0:
v.dirSwapTime /= v.dirSwaps
print "\n"
def calcDevs(vals):
for name in vals:
v = vals[name]
print name+":"
v.szVar /= v.totalOps
v.skVar /= v.totalOps
v.dlVar /= v.totalOps
v.szDev = math.sqrt(v.szVar)
v.skDev = math.sqrt(v.skVar)
v.dlDev = math.sqrt(v.dlVar)
print "size stddev = "+str(v.szDev)
print "seek stddev = "+str(v.skDev)
print "delay stddev = "+str(v.dlDev)
print ""
print "\n"
def calcChi2s(vals):
for name in vals:
v = vals[name]
print name+":"
v.calcChi2()
def writeJobFile(vals):
print "; job file based on "+infilename
print "[global]"
print "filename=FILENAME"
print "direct=1"
print "runtime=60" # 1 minute
print "size=1M"
print "time_based"
print "do_verify=0"
print "zero_buffers"
print "norandommap"
print ""
for name in vals:
if name == "__overall":
continue
v = vals[name]
if v.totalOps < 1:
continue
print "["+name+"]"
print "prioclass=2" #best-effort
print "prio=7" #lowest, 0 is highest
if v.sync >= v.async:
print "ioengine=vsync"
else:
print "ioengine=libaio"
if v.maxDepth > 64:
v.maxDepth = 64
print "iodepth="+str(int(round(v.maxDepth)))
bsrange = ""
if v.reads >= 1:
bsrange += str(v.minRead)+"-"+str(v.maxRead)
if v.writes >= 1:
bsrange += ","
if v.writes >= 1:
bsrange += str(v.minWrite)+"-"+str(v.maxWrite)
print "bsrange="+bsrange
if v.meanThink > 60:
v.meanThink = 60
micros = int(round(v.meanThink * 1000000))
print "thinktime="+str(micros)
meanSequential = int(round(float(v.totalOps) / (v.totalOps - v.sequentialCount)))
if v.reads == 0:
print "rw=randwrite:"+str(meanSequential)
elif v.writes == 0:
print "rw=randread:"+str(meanSequential)
else:
print "rw=randrw:"+str(meanSequential)
if v.reads != 0 and v.writes != 0:
pctread = (float(v.reads) / v.totalOps) * 100
print "rwmixread="+str(int(round(pctread)))
print "rwmixcycle="+str(int(round(v.dirSwapTime)))
print ""
if __name__ == "__main__":
parseArgs()
vals = {}
vals["__overall"] = TSVals() # HACK: overall values (avoid processes named "__overall"...)
infile = open(infilename)
for line in infile:
words = line.split()
thisTime = float(words[4])
name = words[5]
if words[0] == 'C':
handleCompletion(vals, name, thisTime)
continue
if words[0] == 'M' or words[0] == 'F':
handleMerge(vals, name)
continue
rw = words[1][0] == 'W'
sync = words[1].find('S') != -1
size = int(words[2])
sector = int(words[3])
updateRun1(vals, rw, sync, size, sector, thisTime, name)
if bigmem or do_fio:
infile.close()
calcMeans(vals)
if do_fio:
writeJobFile(vals)
else:
for v in vals.values():
v.initBins()
#print "bin widths "+str(szBinWidth)+","+str(skBinWidth)+","+str(dlBinWidth)
if bigmem:
for v in vals.values():
if v.totalOps < 1:
continue
for (rw,size,seek,delay) in v.ops:
v.szVar += (size - v.meanSize)**2
v.szBins[int((size-v.minSize)/v.szBinWidth)] += 1
v.skVar += (seek - v.meanSeek)**2
v.skBins[int((seek-v.minSeek)/v.skBinWidth)] += 1
v.dlVar += (delay - v.meanDelay)**2
v.dlBins[int((delay-v.minDelay)/v.dlBinWidth)] += 1
v.ops = None
else:
for v in vals.values():
v.lastTime = 0.0
v.lastSector = 0
infile.seek(0)
for line in infile:
words = line.split()
if words[0] == 'C' or words[0] == 'M' \
or words[0] == 'F':
continue
size = int(words[2])
sector = int(words[3])
thisTime = float(words[4])
name = words[5]
updateRun2(name, vals, size, sector, thisTime)
infile.close()
calcDevs(vals)
calcChi2s(vals)