-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathMain_Zipf_GammaVar.py
163 lines (118 loc) · 5.55 KB
/
Main_Zipf_GammaVar.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
# This is the main file to run the simulator for our coded load balancing scheme.
#
# Here we derive "maximum load" and "average comm. cost" for Zipf
# distribution versus Zipf parameter gamma (0 \le gamma \le infty).
#
#
from __future__ import division
import math
from multiprocessing import Pool
import sys
import numpy as np
import scipy.io as sio
import time
#from CodedLoadBalancing.Simulator import *
from CodedLoadBalancing.Simulator_MltChnk import *
# --------------------------------------------------------------------
# Simulation parameters
# Choose the simulator. It can be the following values:
simulator = 'Coded'
# Base part of the output file name
base_out_filename = 'ZipfGammaVar'
# Pool size for parallel processing
pool_size = 4
# Total number of runs for computing the average values.
# It is more efficient that num_of_runs be a multiple of pool_size
num_of_runs = 10
# Number of servers
srv_num = 1024
# Number of requests
req_num = srv_num
# Cache size of each server (expressed in number of files)
cache_sz = 2
# Total number of files in the system
file_num = 100
# The number of chunks
# If the number of chunks is set to one, we in fact simulate the nearest replica strategy.
chnk_num = 1
# The maximum number of chunks that can be downloaded from each server.
chnk_max = 1
# The graph structure of the network
# It can be:
# 'Lattice' for square lattice graph. For the lattice the graph size should be perfect square.
# 'RGG' for random geometric graph. The RGG is generate over a unit square or unit cube.
# 'BarabasiAlbert' for Barabasi Albert random graph. It takes two parameters: # of nodes and # of edges
# to attach from a new node to existing nodes
graph_type = 'Lattice'
#graph_type = 'RGG'
#graph_type = 'BarabasiAlbert'
# The parameters of the selected random graph
# The dictionary graph_param should always be defined. However, for some graphs it may not be used.
graph_param = {'rgg_radius' : sqrt(5 / 4 * log(srv_num)) / sqrt(srv_num)} # RGG radius for random geometric graph.
graph_param = {'num_edges' : 1} # For Barbasi Albert random graphs.
# The distribution of file placement in nodes' caches
# It can be:
# 'Uniform' for uniform placement, or
# 'Zipf' for Zipf placement. We have to determine the parameter 'gamma' for this distribution in 'place_dist_param'.
# placement_dist = 'Uniform'
# But here we have to choose Zipf!
placement_dist = 'Zipf'
# The list of Zipf parameters for simulation
gamma_range = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0]
# The parameters of the placement distribution
#place_dist_param = {'gamma': 1.0} # For Zipf distribution where 0 < gamma < infty
# --------------------------------------------------------------------
if __name__ == '__main__':
# Create a number of workers for parallel processing
pool = Pool(processes=pool_size)
t_start = time.time()
rslt_maxload = np.zeros((len(gamma_range), 1 + num_of_runs))
rslt_avgcost = np.zeros((len(gamma_range), 1 + num_of_runs))
rslt_outage = np.zeros((len(gamma_range), 1 + num_of_runs))
for i, gm in enumerate(gamma_range):
place_dist_param = {'gamma' : gm}
params = [(srv_num, req_num, cache_sz, file_num, chnk_num, chnk_max, graph_type, graph_param, placement_dist, place_dist_param)
for itr in range(num_of_runs)]
print(params)
if simulator == 'Coded':
rslts = pool.map(coded_load_balancing_simulator, params)
# rslts = map(Simulator1, params)
# elif simulator == 'TwoChoice':
# rslts = pool.map(simulator_twochoice, params)
else:
print('Error: an invalid simulator!')
sys.exit()
for j, rslt in enumerate(rslts):
rslt_maxload[i, j + 1] = rslt['maxload']
rslt_avgcost[i, j + 1] = rslt['avgcost']
rslt_outage[i, j + 1] = rslt['outage']
# tmpmxld = tmpmxld + rslt['maxload']
# tmpavgcst = tmpavgcst + rslt['avgcost']
rslt_maxload[i, 0] = gm
rslt_avgcost[i, 0] = gm
rslt_outage[i, 0] = gm
# result[i,1] = tmpmxld/num_of_runs
# result[i,2] = tmpavgcst/num_of_runs
t_end = time.time()
print("The runtime is {}".format(t_end - t_start))
# Write the results to a matlab .mat file
if placement_dist == 'Uniform':
sio.savemat(base_out_filename + '_{}_{}_{}_sn={}_fn={}_cs={}_chn={}_chnmax={}_itr={}.mat'.\
format(graph_type, placement_dist, simulator, srv_num, file_num, cache_sz, chnk_num, chnk_max, num_of_runs), \
{'maxload': rslt_maxload, 'avgcost': rslt_avgcost, 'outage': rslt_outage})
elif placement_dist == 'Zipf':
sio.savemat(base_out_filename + '_{}_{}_{}_sn={}_fn={}_cs={}_chn={}_chnmax={}_itr={}.mat'.\
format(graph_type, placement_dist, simulator, srv_num, file_num, cache_sz, chnk_num, chnk_max, num_of_runs), \
{'maxload': rslt_maxload, 'avgcost': rslt_avgcost, 'outage': rslt_outage})
# plt.plot(result[:,0], result[:,1])
# plt.plot(result[:,0], result[:,2])
# plt.xlabel('Cache size in each server (# of files)')
# plt.title('Random server selection methods. Number of servers = {}, number of files = {}'.format(srv_num,file_num))
# plt.show()
# print(result['loads'])
# print("------")
# print('The maximum load is {}'.format(result['maxload']))
# print('The average request cost is {0:0}'.format(result['avgcost']))
# fl_lst = [srvs[i].get_files_list() for i in range(n)]
# print(fl_lst)
print(rslt_outage)