forked from khalefa-phd/learned-indexes
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Learned_Storage.py
458 lines (418 loc) · 18.8 KB
/
Learned_Storage.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
# Main file for storage optimization
import pandas as pd
import numpy as np
from Trained_NN import TrainedNN, ParameterPool, set_data_type, AbstractNN
from btree import BTree
from data.create_data import create_data_storage, Distribution
import time, json, math, getopt, sys, gc, csv
# setting
STORE_NUMBER = 100000
BLOCK_SIZE = 100
# data files for existing data
storePath = {
Distribution.RANDOM: "data/random_s.csv",
Distribution.BINOMIAL: "data/binomial_s.csv",
Distribution.POISSON: "data/poisson_s.csv",
Distribution.EXPONENTIAL: "data/exponential_s.csv",
Distribution.NORMAL: "data/normal_s.csv",
Distribution.LOGNORMAL: "data/lognormal_s.csv"
}
# data files for data to be inserted
toStorePath = {
Distribution.RANDOM: "data/random_t.csv",
Distribution.BINOMIAL: "data/binomial_t.csv",
Distribution.POISSON: "data/poisson_t.csv",
Distribution.EXPONENTIAL: "data/exponential_t.csv",
Distribution.NORMAL: "data/normal_t.csv",
Distribution.LOGNORMAL: "data/lognormal_t.csv"
}
# path for write
pathString = {
Distribution.RANDOM: "Random",
Distribution.BINOMIAL: "Binomial",
Distribution.POISSON: "Poisson",
Distribution.EXPONENTIAL: "Exponential",
Distribution.NORMAL: "Normal",
Distribution.LOGNORMAL: "Lognormal"
}
thresholdPool = {
Distribution.RANDOM: [1, 1],
Distribution.EXPONENTIAL: [1, 10000]
}
useThresholdPool = {
Distribution.RANDOM: [True, True],
Distribution.EXPONENTIAL: [True, False],
}
# binary search for data segment
def part_binary_search(data_list, pos_list, key):
start = 0
end = len(pos_list) - 1
mid = 0
while start <= end:
mid = (start + end) / 2
if data_list[pos_list[mid]] < key:
start = mid + 1
elif data_list[pos_list[mid]] > key:
end = mid - 1
else:
return mid
if data_list[pos_list[mid]] > key and mid != 0:
return mid - 1
else:
return mid
# binary search for position in data segment
def pos_binary_search(data_list, key):
start = 0
end = len(data_list) - 1
mid = 0
while start <= end:
mid = (start + end) / 2
if data_list[mid] == key or data_list[mid] == -1:
return mid
elif data_list[mid] < key:
start = mid + 1
else:
end = mid - 1
if data_list[mid] > key:
return mid - 1
else:
return mid
# function for train index
def hybrid_training(threshold, use_threshold, stage_nums, core_nums, train_step_nums, batch_size_nums,
learning_rate_nums, keep_ratio_nums, train_data_x, train_data_y, test_data_x, test_data_y):
stage_length = len(stage_nums)
col_num = stage_nums[1]
tmp_inputs = [[[] for i in range(col_num)] for i in range(stage_length)]
tmp_labels = [[[] for i in range(col_num)] for i in range(stage_length)]
index = [[None for i in range(col_num)] for i in range(stage_length)]
tmp_inputs[0][0] = train_data_x
tmp_labels[0][0] = train_data_y
test_inputs = test_data_x
for i in range(0, stage_length):
for j in range(0, stage_nums[i]):
if len(tmp_labels[i][j]) == 0:
continue
inputs = tmp_inputs[i][j]
labels = []
test_labels = []
if i == 0:
divisor = stage_nums[i + 1] * 1.0 / (STORE_NUMBER / BLOCK_SIZE)
for k in tmp_labels[i][j]:
labels.append(int(k * divisor))
for k in test_data_y:
test_labels.append(int(k * divisor))
else:
labels = tmp_labels[i][j]
test_labels = test_data_y
tmp_index = TrainedNN(threshold[i], use_threshold[i], core_nums[i], train_step_nums[i], batch_size_nums[i],
learning_rate_nums[i],
keep_ratio_nums[i], inputs, labels, test_inputs, test_labels)
tmp_index.train()
index[i][j] = AbstractNN(tmp_index.get_weights(), tmp_index.get_bias(), core_nums[i], tmp_index.cal_err())
del tmp_index
gc.collect()
if i < stage_length - 1:
for ind in range(len(tmp_inputs[i][j])):
p = index[i][j].predict(tmp_inputs[i][j][ind])
if p > stage_nums[i + 1] - 1:
p = stage_nums[i + 1] - 1
tmp_inputs[i + 1][p].append(tmp_inputs[i][j][ind])
tmp_labels[i + 1][p].append(tmp_labels[i][j][ind])
for i in range(stage_nums[stage_length - 1]):
if index[stage_length - 1][i] is None:
continue
mean_abs_err = index[stage_length - 1][i].mean_err
if mean_abs_err > threshold[stage_length - 1]:
print("Using BTree")
index[stage_length - 1][i] = BTree(2)
index[stage_length - 1][i].build(tmp_inputs[stage_length - 1][i], tmp_labels[stage_length - 1][i])
return index
# calculate data distribution
def learn_density(threshold, use_threshold, distribution, train_set_x, train_set_y, test_set_x, test_set_y):
set_data_type(distribution)
if distribution == Distribution.RANDOM:
parameter = ParameterPool.RANDOM.value
elif distribution == Distribution.LOGNORMAL:
parameter = ParameterPool.LOGNORMAL.value
elif distribution == Distribution.EXPONENTIAL:
parameter = ParameterPool.EXPONENTIAL.value
elif distribution == Distribution.NORMAL:
parameter = ParameterPool.NORMAL.value
else:
return
stage_set = parameter.stage_set
stage_set[1] = int(STORE_NUMBER / 10000)
core_set = parameter.core_set
train_step_set = parameter.train_step_set
batch_size_set = parameter.batch_size_set
learning_rate_set = parameter.learning_rate_set
keep_ratio_set = parameter.keep_ratio_set
print("*************start Learned NN************")
print("Start Train")
# train NN index
start_time = time.time()
trained_index = hybrid_training(threshold, use_threshold, stage_set, core_set, train_step_set, batch_size_set,
learning_rate_set,
keep_ratio_set, train_set_x, train_set_y, test_set_x, test_set_y)
end_time = time.time()
learn_time = end_time - start_time
print("Build Learned NN time %f " % learn_time)
print("*************end Learned NN************")
return trained_index
# main function for storage optimization
def optimize_storage(do_compare, do_record, threshold, use_threshold, data_part_distance, learning_percent, distribution):
store_path = storePath[distribution]
to_store_path = toStorePath[distribution]
tmp_data = pd.read_csv(store_path, header=None)
train_set_x = []
train_set_y = []
test_set_x = []
test_set_y = []
global STORE_NUMBER
STORE_NUMBER = tmp_data.shape[0]
for i in range(STORE_NUMBER):
# test_set_x.append(tmp_data.ix[i, 0])
# test_set_y.append(tmp_data.ix[i, 1])
train_set_x.append(tmp_data.ix[i, 0])
train_set_y.append(tmp_data.ix[i, 1])
store_data = train_set_x[:]
to_store_data = pd.read_csv(to_store_path, header=None)
if do_compare == 1 or do_compare == 2:
trained_index = learn_density(threshold, use_threshold, distribution, train_set_x, train_set_y, test_set_x,
test_set_y)
print("************Start Optimization**************")
stage_size = int(STORE_NUMBER / 10000)
min_value = train_set_x[0]
max_value = train_set_x[-1]
data_density = []
data_density_pos = [0]
# get data segment number according to distance
data_part_num = int(math.ceil((max_value - min_value) * 1.0 / data_part_distance))
last_pre = 0
store_data_num = len(store_data)
# calculate how many blocks have been occupied
store_block_num = int(math.ceil(store_data_num * 1.0 / BLOCK_SIZE))
start_time = time.time()
for i in range(1, data_part_num):
# calculate first data of every data segment
pre_data = min_value + i * data_part_distance
# calculate position of data
pre1 = trained_index[0][0].predict(pre_data)
if pre1 > stage_size - 1:
pre1 = stage_size - 1
pre2 = trained_index[1][pre1].predict(pre_data)
if pre2 > store_block_num:
pre2 = store_block_num
if pre2 <= last_pre:
continue
if pre2 >= store_block_num - 1:
break
# calculate data ratio, using occupied blocks
data_density_pos.append(pre2 * BLOCK_SIZE)
data_density.append(abs(pre2 - last_pre) * 1.0 / store_block_num)
last_pre = pre2
data_density_pos.append(store_data_num)
data_density.append(abs(store_block_num - 1 - last_pre) * 1.0 / store_block_num)
data_part_num = len(data_density)
# enlarge storage according to estimation
store_data = train_set_x[:]
total_data_num = int(math.ceil(store_block_num * BLOCK_SIZE * (1.0 / learning_percent)))
for i in range(total_data_num - store_data_num):
store_data.append(-1)
block_pos = total_data_num - int(
math.ceil(total_data_num * (1.0 / store_block_num)))
data_optimization_pos = []
data_free_pos = []
# move data to new position
for i in range(data_part_num, 0, -1):
block_pos -= int(round(data_density[i - 1] * total_data_num))
if block_pos <= 0:
data_optimization_pos.insert(0, 0)
data_free_pos.insert(0, data_density_pos[i])
break
# record first data position in every data segment
data_optimization_pos.insert(0, block_pos)
# move data to new position
store_data[block_pos: block_pos + data_density_pos[i] - data_density_pos[i - 1]] = \
store_data[data_density_pos[i - 1]:data_density_pos[i]]
# free space in old position
if block_pos < data_density_pos[i]:
store_data[data_density_pos[i - 1]:block_pos] = [-1] * (block_pos - data_density_pos[i - 1])
else:
store_data[data_density_pos[i - 1]:data_density_pos[i]] = [-1] * (
data_density_pos[i] - data_density_pos[i - 1])
# record first free position (for insertion) in every data segment
data_free_pos.insert(0, block_pos + data_density_pos[i] - data_density_pos[i - 1])
end_time = time.time()
average_optimize_time = (end_time - start_time) * 1.0 / to_store_data.shape[0]
print("Average Optimize Time: %lf" % average_optimize_time)
std_deviation = np.std(data_density)
mean_density = np.mean(data_density)
print("Density Standard Deviation: %f" % std_deviation)
print("Mean Density: %f" % mean_density)
if do_record:
with open('optimization_result.csv', 'wb') as csvFile:
csv_writer = csv.writer(csvFile)
for i in store_data:
csv_writer.writerow([i])
move_steps = len(train_set_x)
# test optimization
print("************With Optimization**************")
start_time = time.time()
for i in range(to_store_data.shape[0]):
pre_data = to_store_data.ix[i, 0]
# calculate isertion position
# find data segment for insertion
part = part_binary_search(store_data, data_optimization_pos, pre_data)
# find position in data segment for insertion
pos = data_optimization_pos[part] + pos_binary_search(
store_data[data_optimization_pos[part]: data_free_pos[part]], pre_data)
ins_pos = data_free_pos[part]
# insert data
while store_data[ins_pos] != -1 and ins_pos < len(store_data) - 1:
ins_pos += 1
if ins_pos == len(store_data) - 1:
store_data.append(-1)
store_data[pos + 2: ins_pos + 1] = store_data[pos + 1:ins_pos]
data_free_pos[part] = ins_pos + 1
store_data[pos + 1] = pre_data
move_steps += ins_pos - pos
end_time = time.time()
# calculate moving steps and time
average_move_steps = (move_steps * 1.0 / to_store_data.shape[0])
average_move_time = (end_time - start_time) * 1.0 / to_store_data.shape[0]
average_insert_time = average_move_time + average_optimize_time
print("Average Move Steps: %f" % average_move_steps)
print("Average Move Time: %f" % average_move_time)
print("Average Insert Time: %f" % average_insert_time)
result = [{"Average Moving Steps": average_move_steps, "Average Moving Time": average_move_time,
"Average Optimizing Time": average_optimize_time, "Average Insert Time": average_insert_time,
"Mean Density": mean_density, "Density Standard Deviation": std_deviation}]
with open("store_performance/" + pathString[distribution] + "/optimization/" + str(
data_part_distance) + "_" + str(learning_percent) + ".json", "wb") as jsonFile:
json.dump(result, jsonFile)
if do_record:
with open('insert_result.csv', 'wb') as csvFile:
csv_writer = csv.writer(csvFile)
for i in store_data:
csv_writer.writerow([i])
if do_compare == 0 or do_compare == 2:
# test no optimization
print("************Without Optimization**************")
store_data = train_set_x[:]
move_steps = 0
start_time = time.time()
for i in range(to_store_data.shape[0]):
pre_data = to_store_data.ix[i, 0]
pos = pos_binary_search(store_data, pre_data)
store_data.append(-1)
store_data[pos + 2:len(store_data)] = store_data[pos + 1:len(store_data) - 1]
store_data[pos + 1] = pre_data
move_steps += len(store_data) - pos - 3
end_time = time.time()
average_move_steps = (move_steps * 1.0 / to_store_data.shape[0])
average_move_time = (end_time - start_time) * 1.0 / to_store_data.shape[0]
print("Average Move Steps: %f" % average_move_steps)
print("Average Move Time: %f" % average_move_time)
result = [{"Average Moving Steps": average_move_steps, "Average Moving Time": average_move_time}]
with open("store_performance/" + pathString[distribution] + "/no_optimization/"
+ str(learning_percent) + ".json", "wb") as jsonFile:
json.dump(result, jsonFile)
# help message
def show_help_message(msg):
help_message = {
'command': 'python Learned_BTree.py -d <Distribution> [-p] [Percent] '
'[-s] [Distance] [-c] [Compare] [-n] [New data] [-r] [Record] [-h]',
'distribution': 'Distribution: random, exponential',
'percent': 'Percent: 0.1-1.0, default value = 0.5; train data size = 100,000',
'distance': 'Distance:'
'[Random: 100-100,000, default = 1,000; '
'Exponential: 100,000-100,000,000, default = 1,000,000]',
'compare': 'Compare: INTEGER, 2 for comparing, 1 for only optimization, 0 for only no optimization, default = 2',
'new data': 'New data: INTEGER, 0 for no creating new data file, others for creating, default = 1',
'record': 'Record: INTEGER, 0 for no printing out result, others for printing, default = 0',
'noDistributionError': 'Please choose the distribution first.'}
help_message_key = ['command', 'distribution', 'percent', 'distance', 'compare', 'new data', 'record']
if msg == 'all':
for k in help_message_key:
print(help_message[k])
else:
print(help_message['command'])
print('Error! ' + help_message[msg])
# command line
def main(argv):
distribution = None
per = 0.5
num = 100000
is_distribution = False
distance = 1000
do_compare = 2
do_create = True
do_record = False
try:
opts, [] = getopt.getopt(argv, "hd:s:p:c:n:r:")
except getopt.GetoptError:
show_help_message('command')
sys.exit(2)
for opt, arg in opts:
arg = str(arg).lower()
if opt == '-h':
show_help_message('all')
return
elif opt == '-d':
if arg == "random":
distribution = Distribution.RANDOM
is_distribution = True
distance = 1000
elif arg == "exponential":
distribution = Distribution.EXPONENTIAL
is_distribution = True
distance = 1000000
else:
show_help_message('distribution')
return
elif opt == '-p':
if not is_distribution:
show_help_message('noDistributionError')
return
per = float(arg)
if not 0.1 <= per <= 1.0:
show_help_message('percent')
return
elif opt == '-s':
if not is_distribution:
show_help_message('noDistributionError')
return
distance = int(arg)
if not 10 <= distance <= 100000000:
show_help_message('distance')
return
elif opt == '-c':
if not is_distribution:
show_help_message('noDistributionError')
return
do_compare = int(arg)
if not (do_compare == 0 or do_compare == 1 or do_compare == 2):
return
elif opt == '-n':
if not is_distribution:
show_help_message('noDistributionError')
return
do_create = not (int(arg) == 0)
elif opt == '-r':
if not is_distribution:
show_help_message('noDistributionError')
return
do_record = not (int(arg) == 0)
else:
print("Unknown parameters, please use -h for instructions.")
return
if not is_distribution:
show_help_message('noDistributionError')
return
if do_create:
create_data_storage(distribution, per, num)
optimize_storage(do_compare, do_record, thresholdPool[distribution], useThresholdPool[distribution], distance, per,
distribution)
if __name__ == "__main__":
main(sys.argv[1:])