-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathwiki_workload_generator_complex_SFC.R
481 lines (385 loc) · 15.4 KB
/
wiki_workload_generator_complex_SFC.R
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
# This workload generator creates a workload for SFC experiment.
# Arrival rate: following the Wikipedia trace.
# Network worklaod: each pakcet has the same constant packet size, to simplify the network simulation
# CPU workload: each workload has the same CPU size.
# We assumed that Load Balancer already worked, so that the workload is assigned in round-robin.
# Two kinds of workoad
# 1. Cached workload which only arrives at Web server, no access to App or DB.
# This implies the Web server had a cached data.
# 2. Traditional worklaod that passes Web, App, and DB servers.
library(actuar)
library(pracma)
B=1
KB=B*1024
MB=KB*1024
MAX_CPU_MI = 20000 #400 MIs = 0.2 sec for 2000 MIPS vm
options(scipen=500) #instead of E notation, use full representation.
# Parameters for history based workloads
NETWORK_PACKET_SIZE_MULTIPLY = 1
CPU_MULTIPLY = 1
NUM_WORKLOAD_RATE_SCALE = 1
#average FRONT(web) MID(app) BACK(bd)
PACKETSIZE_FRONT_MID = 1500
PACKETSIZE_MID_BACK = 1500
PACKETSIZE_BACK_MID = 65000
PACKETSIZE_MID_FRONT = 65000
#PACKETSIZE_FRONT_MID = 500
#PACKETSIZE_MID_BACK = 500
#PACKETSIZE_BACK_MID = 22800
#PACKETSIZE_MID_FRONT = 345800
CPUSIZE_FRONT_1 = 72*CPU_MULTIPLY
CPUSIZE_MID_1 = 585*CPU_MULTIPLY
CPUSIZE_BACK_1 = 306*CPU_MULTIPLY
CPUSIZE_MID_2 = 67*CPU_MULTIPLY
CPUSIZE_FRONT_2 = 68*CPU_MULTIPLY
rlnorm2 <-function(n,u,s) {
r = rnorm(n, mean=u, sd=s)
return(2^r)
}
generate_atimes_lnorm<-function(n,u,s, scale=1000) {
r <- rlnorm2(n, u, s)/scale
#cat("median inter-arrival time:", median(r),", throughput:", 1/mean(r),"req/sec\n")
return(round(cumsum(r), digits=4))
}
generate_atimes_dur<-function(u,s, target.rate = -1, dur=1, scale=1) {
n = target.rate*dur*1.5
times = generate_atimes_lnorm(n,u,s,scale)
if(target.rate != -1) {
# Spread out the time.
throughput = n / tail(times,n=1)
ratio = throughput / target.rate
times = times * ratio
}
times=times[times<=dur]
#cat("Adjusted throughput=", n / tail(times,n=1),"req/sec\n")
return(round(times, digits=4))
}
generate_repeat<-function(n,mean) {
sd = mean*0.2
r <- rnorm(n, mean, sd)
return(round(r))
}
generate_packet_size<-function(n,u,s, scale) {
r <- rlnorm2(n, u, s) * scale
cat("median packet size:", median(r), "\n")
return(round(r))
}
generate_cpu_size<-function(n, location, shape, vm_cap=1) {
#vm_cap = vm_cap*SCALE_FACTOR
r = rpareto(n*2, shape=shape, scale=location)
r = round(r*vm_cap)
r = r[r<MAX_CPU_MI]
while(length(r) < n) {
rn = rpareto(n, shape=shape, scale=location)
rn = round(rn*vm_cap)
r = c(r, rn[rn<MAX_CPU_MI])
}
r=r[1:n]
#cat("cpu workloads -- median size: ", median(r), ", max:", max(r), ", mean:", mean(r),"\n" )
return(r)
}
combine_works<-function(atime, name.1, name.2, name.3,
w.1.1, p.1.2, link.1.2,
w.2.1, p.2.3, link.2.3,
w.3, p.3.2, link.3.2,
w.2.2, p.2.1, link.2.1,
w.1.2) {
#cat("length of atime=", length(atime), "\n")
#atime=aggregate_first(atime, aggregate_num)
#cat("aggregated to=", length(atime), "\n")
"
w.1.1=aggregate_sum(w.1.1, aggregate_num)
p.1.2=aggregate_sum(p.1.2, aggregate_num)
w.2.1=aggregate_sum(w.2.1, aggregate_num)
p.2.3=aggregate_sum(p.2.3, aggregate_num)
w.3=aggregate_sum(w.3, aggregate_num)
p.3.2=aggregate_sum(p.3.2, aggregate_num)
w.2.2=aggregate_sum(w.2.2, aggregate_num)
p.2.1=aggregate_sum(p.2.1, aggregate_num)
w.1.2=aggregate_sum(w.1.2, aggregate_num)
"
zeros= rep(0, length(atime))
wl.0.1=cbind( name.1, zeros, w.1.1)
wl.1.2=cbind(link.1.2, name.2, p.1.2, w.2.1)
wl.2.3=cbind(link.2.3, name.3, p.2.3, w.3)
wl.3.2=cbind(link.3.2, name.2, p.3.2, w.2.2)
wl.2.1=cbind(link.2.1, name.1, p.2.1, w.1.2)
workload = cbind(atime, wl.0.1, wl.1.2, wl.2.3, wl.3.2, wl.2.1)
return(workload)
}
# This function aggregates the number of workloads into a single workload by summing the aggregating members
aggregate_sum<-function(vec, membernum) {
if(membernum == 1)
return(vec)
n = length(vec)
k = ceiling(n/membernum)
summed = tapply(vec,as.integer(gl(k,membernum, n)),sum)
return(as.vector(summed))
}
# This function aggregates the number of workloads into a single workload by choosing the first member among aggregating member
aggregate_first<-function(vec, membernum) {
n = length(vec)
k = ceiling(n/membernum)
summed = tapply(vec,as.integer(gl(k,membernum, n)),head, n=1)
return(as.vector(summed))
}
create_workload<-function(n = -1, durtime = -1,
target.rate=-1, starttime=0, seed = 10, filename="workloads.csv",
name1set, name2set, name3set,
is.dedicated=FALSE, is.append = FALSE) {
#cat("=== Generating workload.. Start=",starttime,", Target rate =",target.rate, ", n=",n,", dur=",durtime,"\n")
#WebServer tail_index location K-S Test Statistic
#RUBiS, 800 Clients 0.3164 21.364 0.02941
#RUBiS, 1600 Clients 0.5456 22.630 0.03647
#RUBiS, 2400 Clients 0.6423 28.065 0.04529
#RUBiS, 3200 Clients 1.1108 67.729 0.03317
#SPEC, txRate=10 0.6271 24.432 0.06182
#SPEC, txRate=20 0.8202 99.031 0.05742
#SPEC, txRate=40 1.1580 614.98 0.05911
#AppServer tail_index location K-S Test Statistic
#RUBiS, 800 Clients 0.1030 24.6649 0.03554
#RUBiS, 1600 Clients 0.3860 25.2826 0.04398
#RUBiS, 2400 Clients 0.4642 29.8927 0.05172
#RUBiS, 3200 Clients 0.9822 67.8342 0.04439
#SPEC, txRate=10 0.5838 24.4444 0.02021
#SPEC, txRate=20 0.8995 41.2255 0.06374
#SPEC, txRate=40 1.2331 534.159 0.02218
#DBServer tail_index location K-S Test Statistic
#RUBiS, 800 Clients 0.9296 6.7933 0.04681
#RUBiS, 1600 Clients 1.0264 13.1436 0.03266
#RUBiS, 2400 Clients 1.2391 23.3437 0.03541
#RUBiS, 3200 Clients 1.4648 49.2665 0.06184
#SPEC, txRate=10 0.9713 12.3486 0.05922
#SPEC, txRate=20 1.1673 34.6912 0.08214
#SPEC, txRate=40 1.2791 82.5146 0.07734
####################################################
# For SPEC tx=10
t.user.front.mean = 1.56270
t.user.front.sd = 1.5458
#t.front.mid.mean = 2.0815
#t.front.mid.sd = 1.1506
#t.mid.back.mean = 1.0354
#t.mid.back.sd = 0.4157
w.back.location = 12.3486
w.back.shape = 0.9713
#w.front=generate_cpu_size(n,24.432, 0.6271)
#w.mid=generate_cpu_size(n,24.4444,0.5838)
psize.front.mid.mean = 5.6129
psize.front.mid.sd = 0.1343
psize.mid.back.mean = 4.6455
psize.mid.back.sd = 0.8013
psize.back.mid.mean = 3.6839
psize.back.mid.sd = 0.8261
psize.mid.front.mean = 7.0104
psize.mid.front.sd = 0.8481
#psize.front.user.mean = 7.01865
#psize.front.user.sd = 0.4894
####################################################
####################################################
# For RUBIS 3200 client
t.user.front.mean = 0.43154
t.user.front.sd = 1.1225
#t.front.mid = generate_atimes_wei(n, 0.03548, 0.24606)
#t.mid.back.mean = 1.6770
#t.mid.back.sd = 0.8912
w.web.location = 67.729
w.web.shape = 1.1108
w.app.location = 67.8342
w.app.shape = 0.9822
w.db.location = 49.2665
w.db.shape = 1.4648
psize.front.mid.mean = 5.5974
psize.front.mid.sd = 0.1602
psize.mid.back.mean = 4.7053
psize.mid.back.sd = 0.4034
psize.back.mid.mean = 2.2134
psize.back.mid.sd = 0.5126
psize.mid.front.mean = 5.6438
psize.mid.front.sd = 0.5436
#psize.front.user.mean = 5.06502
#psize.front.user.sd = 0.5458
####################################################
w.1.1.location = w.1.2.location = w.web.location
w.2.1.location = w.2.2.location = w.app.location
w.3.location = w.db.location
w.1.1.shape = w.1.2.shape = w.web.shape
w.2.1.shape = w.2.2.shape = w.app.shape
w.3.shape = w.db.shape
scale_factor_net = NETWORK_PACKET_SIZE_MULTIPLY
t.user.front.mean = 0.59481
t.user.front.sd = 1.5284
set.seed(seed)
if( durtime != -1) {
t.user.front = generate_atimes_dur(t.user.front.mean, t.user.front.sd, target.rate/durtime, dur=durtime)
n=length(t.user.front)
} else {
t.user.front = generate_atimes(n, t.user.front.mean, t.user.front.sd, target.rate)
durtime = tail(t.user.front, n=1)
}
#cat("=== Generating workload.. Start=",starttime,", Target rate =",target.rate, ", n=",n,", dur=",durtime,"\n")
if(n == 0L)
return(NA)
t.user.front = t.user.front + starttime
# Option 1: Use a constant size.
psize.front.mid = round(generate_repeat(n, PACKETSIZE_FRONT_MID)*scale_factor_net)
psize.mid.back = round(generate_repeat(n, PACKETSIZE_MID_BACK)*scale_factor_net)
psize.back.mid = round(generate_repeat(n, PACKETSIZE_BACK_MID)*scale_factor_net)
psize.mid.front = round(generate_repeat(n, PACKETSIZE_MID_FRONT)*scale_factor_net)
"
w.1.1 = generate_repeat(n=n, CPUSIZE_FRONT_1)
w.2.1 = generate_repeat(n=n, CPUSIZE_MID_1)
w.3 = generate_repeat(n=n, CPUSIZE_BACK_1)
w.2.2 = generate_repeat(n=n, CPUSIZE_MID_2)
w.1.2 = generate_repeat(n=n, CPUSIZE_FRONT_2)
"
# Option 2: Generate new workloads
"
set.seed(seed+1)
psize.front.mid = generate_packet_size(n, psize.front.mid.mean, psize.front.mid.sd, KB)*scale_factor_net
set.seed(seed+2)
psize.mid.back = generate_packet_size(n, psize.mid.back.mean, psize.mid.back.sd, KB)*scale_factor_net
set.seed(seed+3)
psize.back.mid = generate_packet_size(n, psize.back.mid.mean, psize.back.mid.sd, KB)*scale_factor_net
set.seed(seed+4)
psize.mid.front = generate_packet_size(n, psize.mid.front.mean, psize.mid.front.sd, KB)*scale_factor_net
#psize.front.user = generate_packet_size(n, psize.front.user.mean, psize.front.user.sd, KB)
"
set.seed(seed+5)
w.1.1=generate_cpu_size(n=n, location=w.1.1.location, shape=w.1.1.shape)
set.seed(seed+6)
w.1.2=generate_cpu_size(n=n, location=w.1.2.location, shape=w.1.2.shape)
set.seed(seed+7)
w.2.1=generate_cpu_size(n=n, location=w.2.1.location, shape=w.2.1.shape)
set.seed(seed+8)
w.2.2=generate_cpu_size(n=n, location=w.2.2.location, shape=w.2.2.shape)
set.seed(seed+9)
w.3=generate_cpu_size(n=n, location=w.3.location, shape=w.3.shape)
# Prepare server names (webX-Y, appX-Y, dbX-Y)
name1=rep_len(name1set, n)
name2=rep_len(name2set, n)
name3=rep_len(name3set, n)
link.1.2 = link.2.3 = link.3.2 = link.2.1 = "default"
if(is.dedicated) {
link.1.2 = paste(name1,name2, sep="")
link.2.3 = paste(name2,name3, sep="")
link.3.2 = paste(name3,name2, sep="")
link.2.1 = paste(name2,name1, sep="")
}
workload= combine_works(t.user.front, name1, name2, name3,
w.1.1, psize.front.mid, link.1.2,
w.2.1, psize.mid.back, link.2.3,
w.3, psize.back.mid, link.3.2,
w.2.2, psize.mid.front, link.2.1,
w.1.2)
#write.table(workload, file=filename, row.names=FALSE, quote = FALSE, sep=",", col.names = !is.append, append = is.append)
#return(length(t.user.front)) # returns median of the first CPU MIs
return(workload)
}
is.list.empty<-function(workload_list, index) {
return(is.na(workload_list[[index]][1]))
}
create_historical_data<-function(langId, timelist, ratelist, webNum, appNum, dbNum,
startTime=0, endTime=.Machine$integer.max,
use.dedicated.network = TRUE, fileId=-1
) {
if(fileId == -1)
fileId = langId
is_append=FALSE
interevent = timelist[2] - timelist[1]
maxNum = Lcm(webNum, Lcm(appNum, dbNum))
webSeq = rep_len(seq(0,webNum-1), maxNum)
appSeq = rep_len(seq(0,appNum-1), maxNum)
dbSeq = rep_len(seq(0,dbNum-1), maxNum)
webNamePool = paste(sprintf("web%d-", langId), webSeq, sep="")
appNamePool = paste(sprintf("app%d-", langId), appSeq, sep="")
dbNamePool = paste(sprintf("db%d-", langId), dbSeq, sep="")
workload_list = list() # List to store all workloads
for(workloadId in 1:(maxNum)) {
workload_list[[workloadId]] = NA
}
for(i in 1:(length(ratelist))) {
stime = timelist[i]
if( (stime < startTime) || (stime > endTime)) {
next
}
stime = stime - startTime
rate_per_vm = NUM_WORKLOAD_RATE_SCALE*ratelist[i] / maxNum
for(workloadId in 1:(maxNum)) {
seed = langId*31+i*13+(workloadId-1)*7+127
webName = webNamePool[workloadId]
appName = appNamePool[workloadId]
dbName = dbNamePool[workloadId]
workload = create_workload(durtime=interevent, n=-1,
seed=seed, target.rate = rate_per_vm, starttime=stime,
name1set=webName, name2set=appName, name3set=dbName,
filename=NA , is.dedicated=use.dedicated.network, is.append=is_append)
# Store workload in the list. for each workload ID.
if(!is.na(workload[1])) {
if(!is.list.empty(workload_list, workloadId))
workload_list[[workloadId]] = rbind(workload_list[[workloadId]], workload)
else
workload_list[[workloadId]] = workload
}
}
# every 300 rounds, flush workload_list to a file
if(i %% 100 == 0) {
cat("=== Saving workload.. Start=",stime,", Target rate =",rate_per_vm, ", dur=",interevent,"\n")
for(workloadId in 1:(maxNum)) {
if(!is.list.empty(workload_list, workloadId)) {
filename = sprintf("%d_%d_workload_wiki.csv", langId, workloadId-1)
write.table(workload_list[[workloadId]], file=filename, row.names=FALSE, quote = FALSE, sep=",", col.names = !is_append, append = is_append)
workload_list[[workloadId]] = NA
}
}
is_append=TRUE
}
}
for(workloadId in 1:(maxNum)) {
if(!is.list.empty(workload_list, workloadId)) {
filename = sprintf("%d_%d_workload_wiki.csv", langId, workloadId-1)
write.table(workload_list[[workloadId]], file=filename, row.names=FALSE, quote = FALSE, sep=",", col.names = !is_append, append = is_append)
workload_list[[workloadId]] = NA
}
}
}
#endTime=.Machine$integer.max
main_history<-function(lang="de", startTime=0, endTime=5000) {
#history_file="overallstatpersec9-1-mod3.csv"
history_file="overallstatpersec9-1-sfc.csv"
historyData = read.csv(file=history_file,sep=',',header=F)
if(lang != 0) {
# pass lang either "de", "fr", "ru", "zh", or "es"
"
langId=0
fileId=1
create_historical_data(langId, timelist=historyData$V1[historyData$V2==lang],
ratelist=historyData$V3[historyData$V2==lang],
webNum=8,
appNum=0,
dbNum=0,
startTime=startTime, endTime=endTime, fileId=fileId)
"
langId=0
fileId=0
lang="de"
create_historical_data(langId, timelist=historyData$V1[historyData$V2==lang],
ratelist=historyData$V3[historyData$V2==lang],
webNum=40,
appNum=120,
dbNum=10,
startTime=startTime, endTime=endTime, fileId=fileId)
}
else {
# create workload for each language set
i=0
for(lang in levels(historyData$V2)) {
create_historical_data(i, timelist=histoinsinryData$V1[historyData$V2==lang],
ratelist=historyData$V3[historyData$V2==lang],
webNum=8,
appNum=24,
dbNum=2,
startTime=startTime, endTime=endTime)
i=i+1
}
}
}