-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathconfig.py
235 lines (190 loc) · 9.99 KB
/
config.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
import yaml
from job import *
import pandas as pd
import numpy as np
import os, pickle
import sys
from util import Event, Request
from client import *
from src.venn import *
from src.venn_job import *
from src.venn_client import *
setup_file = str(sys.argv[5]) if len(sys.argv) > 5 else None
config = None
if setup_file:
with open(setup_file, "r") as yamlfile:
config = yaml.load(yamlfile, Loader=yaml.FullLoader)
print("Read successful: ", config)
num_job = NUM_JOB = int(sys.argv[4]) if len(sys.argv)>4 else 2
RANDOMSEED = config['RANDOMSEED'] if config else 102
NUM_DAY = config['NUM_DAY'] if config else 5
NUM_WEEK = config['NUM_WEEK'] if config else 4
client_file = config['client_file'] if config else 'trace/fedscale_clients_7000000_3_info.csv'
eligibility_file = config['eligibility_file'] if config else 'trace/baseline_eligibility_3skew'
random.seed(RANDOMSEED)
np.random.seed(RANDOMSEED)
apple_job = ['AppleJob', 'DecAppleJob']
google_job = ['Job', 'AgnosticJob', 'GoogleJob', 'DecentralizedJob', 'VennGoogleJob']
async_job = ['PapayaJob', 'DecPapayaJob', 'VennPapayaJob']
venn_job_name = ['VennGoogleJob', 'VennAppleJob', 'VennPapayaJob' ]
def load_device_capacity(file_path = 'trace/client_device_capacity'):
global_client_profile = {}
if os.path.exists(file_path):
with open(file_path, 'rb') as fin:
# {clientId: [computer, bandwidth]}
global_client_profile = pickle.load(fin)
return global_client_profile
def load_device_eligibility(eligibility_file):
with open(eligibility_file, 'rb') as config_dictionary_file:
eligibility = pickle.load(config_dictionary_file)
return eligibility
def load_device_state(client_file, eligibility_file, client_type , days ):
client_capacity = load_device_capacity()
client_eligibility = load_device_eligibility(eligibility_file)
num_cap = len(client_capacity)
num_eli = len(client_eligibility)
avg_comp = 78
avg_comm = 13736
weeks = NUM_WEEK
client_event_list= []
client_trace = pd.read_csv(client_file)
client_type = eval(client_type)
print(f"Using {client_type}")
for i, (ind, client) in enumerate(client_trace.sort_values(by=['start']).iterrows()):
# for i, (ind, client) in enumerate(client_trace.iterrows() ):
if client['start'] > days * 86400:
break
# if client['start'] < 120: # or client['end'] - client['start'] < 20:
# continue
comm = client_capacity[i % num_cap + 1]['communication'] / avg_comm
comp = client_capacity[i % num_cap + 1]['computation'] / avg_comp
# c = client_type(i, client['start'], client['end'], comp, comm, client_eligibility[i%num_eli])
# client_event_list.append( Event(client['start'], 'CHECKIN', i, c) )
for j in range(weeks):
# TODO: client online period is too short; remove some short-lived clients
# while manually increase their online period to ensure enough traffic
start = client['start'] + j * 432000
c = client_type( i*weeks+j, start, client['end'] + 80 + j * 432000, comp, comm, client_eligibility[i % num_eli])
client_event_list.append(Event(start, 'CHECKIN', i*weeks+j, c))
# c = client_type( -i*weeks+j, client['start'] + j * 432000, client['end'] + 80 + j * 432000, comp, comm, client_eligibility[i % num_eli])
# client_event_list.append(Event(client['start'] + j * 432000, 'CHECKIN', -i*weeks+j, c))
# if i > 10000 :
# break
if i % 100000 == 0:
print(f'Checkin {i*weeks} clients')
return client_event_list
arrival_interval = 1800
start_time = 3600
job_start_list = [start_time] # [0 for _ in range(num_job)]
job_minresponse_list = [0.8 for _ in range(num_job)]
comm_time = 10
timeout_list = [0 for _ in range(num_job)] #
job_workload_list = [60 for _ in range(num_job)] # sample from 30-60?
job_deadline_list = [300 for _ in range(num_job)] #
for i in range(num_job - 1):
job_start_list.append(job_start_list[-1] + np.random.exponential(scale=arrival_interval))
# eligibility_list = np.random.choice([*range(0, 3)], num_job, p=[ 0.33, 0.33 , 0.34])
if not config:
# motivation example
# job_async_round_list = [40000 for _ in range(num_job)] #
buffer_size_list = [10 for _ in range(num_job) ]#
concurrency_list = [100 for _ in range(num_job) ]#
job_round_list = [2000 for _ in range(num_job)] #
job_request_amount_list = [200 for _ in range(num_job)]#
# job_workload_list = np.random.choice(range(30, 61, 10), num_job) # [45 for _ in range(num_job)] #
# job_round_list = np.random.choice(range(5000, 10001, 1000), num_job)
#
# buffer_size_list = np.random.choice(range(10, 31, 5), num_job) # [20 for _ in range(num_job) ]#
# concurrency_list = np.random.choice(range(50, 301, 50), num_job) # [100 for _ in range(num_job) ]#
# timeout_list = job_workload_list + np.random.choice(range(60, 91, 10), num_job)
#
# job_round_list = np.random.choice(range(500, 1001, 100), num_job)
# job_request_amount_list = np.random.choice(range(100, 301, 100), num_job) # [300 for _ in range(num_job)]#
# job_deadline_list = np.random.choice(range(600, 901, 50), num_job) # [300 for _ in range(num_job)] #
def generate_job_by_config(num_job, job_type):
with open('config/job_config.yml', "r") as configfile:
job_config = yaml.load(configfile, Loader=yaml.FullLoader)
job_deadline_list = [0 for _ in range(num_job)] #
request_list = []
job_list = []
job_request_list = []
jobs = job_config['jobs']
job_id_prob = np.array( config['job_id_prob'])
job_type_prob = config['job_type_prob']
job_id_prob =job_id_prob/ sum (job_id_prob)
generate_jobid_by_prob = np.random.choice([*range(len(jobs))], num_job, p = job_id_prob )
print("Random generate job id: ", generate_jobid_by_prob)
jtype = job_type
if jtype in async_job:
job_deadline_list = timeout_list
for i, job_id in enumerate(generate_jobid_by_prob):
print(jobs[job_id])
if job_type == 'MixedJob':
if jobs[job_id]['config']['job_type'] == 'sync':
jtype = np.random.choice(venn_job_name[:2], 1, p = job_type_prob[:2] )[0]
else:
jtype = 'VennPapayaJob'
num_part = jobs[job_id]['config']['participants']
if jtype in google_job:
job_deadline_list[i] = jobs[job_id]['config']['deadline']
# max(min(num_part * 3, 900), 300)
job_request_list += [Request(job_deadline_list[i], num_part,
job_minresponse_list[i], job_workload_list[i], comm_time,
jobs[job_id]['config']['eligibility'] )]
job_list += [eval(jtype)(i, jtype, job_request_list[i], jobs[job_id]['config']['rounds'],
job_start_list[i], jobs[job_id]['config']['concurrency'])]
# concurrency only for async --> is actually buffer size
for job in job_list:
request_list += job.generate_round_event()
request_list += [Event(NUM_DAY * 86400 * NUM_WEEK, 'END', None, None)]
return request_list, job_list, job_request_list
def generate_mixed_job_request(num_job, days):
weeks = NUM_WEEK
request_list = []
job_list = []
job_request_list = []
job_type_list = np.random.choice(venn_job_name, num_job)
for i, job_type in enumerate(job_type_list):
if job_type in async_job:
job_request_list += [Request(timeout_list[i], concurrency_list[i],
job_minresponse_list[i], job_workload_list[i], comm_time,
i % 3)]
job_list += [eval(job_type)(i, job_type, job_request_list[i], job_round_list[i],
job_start_list[i], buffer_size_list[i])]
else:
job_request_list += [Request(job_deadline_list[i], job_request_amount_list[i],
job_minresponse_list[i], job_workload_list[i], comm_time,
i%3 )]
job_list += [eval(job_type)(i, job_type, job_request_list[i], job_round_list[i],
start=job_start_list[i])]
for job in job_list:
request_list += job.generate_round_event()
request_list += [Event(days * 86400 * weeks, 'END', None, None)]
return request_list, job_list, job_request_list
def generate_job_request(num_job, job_type , days = 5):
if job_type == 'RandomJob':
return generate_mixed_job_request(num_job, days)
if config: # MixedJob / config
return generate_job_by_config(num_job, job_type)
weeks = NUM_WEEK
request_list = []
# job_start_list = sorted(np.random.poisson(lam=days/4*10, size=num_job) * 8640)
job_start_list = [start_time] # [0 for _ in range(num_job)]
for i in range(num_job-1):
job_start_list.append( job_start_list[-1] + np.random.exponential(scale=arrival_interval) )
if job_type in async_job:
job_request_list = [Request(timeout_list[i], concurrency_list[i],
job_minresponse_list[i], job_workload_list[i], comm_time,
i%3 ) for i in range(num_job)]
job_list = [eval(job_type)(i, job_type, job_request_list[i], job_round_list[i],
job_start_list[i], buffer_size_list[i]) for i in range(num_job)]
else:
job_request_list = [Request(job_deadline_list[i], job_request_amount_list[i],
job_minresponse_list[i], job_workload_list[i], comm_time,
i%3 ) for i in range(num_job)]
job_list = [eval(job_type)(i, job_type, job_request_list[i], job_round_list[i],
start=job_start_list[i]) for i in range(num_job)]
for job in job_list:
request_list += job.generate_round_event()
request_list += [Event(days * 86400 * weeks, 'END', None, None)]
return request_list, job_list, job_request_list