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
|
#!/usr/bin/env python
#
# Public Domain 2014-2019 MongoDB, Inc.
# Public Domain 2008-2014 WiredTiger, Inc.
#
# This is free and unencumbered software released into the public domain.
#
# Anyone is free to copy, modify, publish, use, compile, sell, or
# distribute this software, either in source code form or as a compiled
# binary, for any purpose, commercial or non-commercial, and by any
# means.
#
# In jurisdictions that recognize copyright laws, the author or authors
# of this software dedicate any and all copyright interest in the
# software to the public domain. We make this dedication for the benefit
# of the public at large and to the detriment of our heirs and
# successors. We intend this dedication to be an overt act of
# relinquishment in perpetuity of all present and future rights to this
# software under copyright law.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
# IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
# ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
# OTHER DEALINGS IN THE SOFTWARE.
# This benchmark is designed to stress disk access to the LAS/History Store file.
# This is achieved through:
# - Long running transactions consisting of read and update operations.
# - Low cache size (~20%) for a reasonably sized WT table with large documents.
# - Pareto distribution for operations in long running transactions. This will cause
# skewed access for a selective set of keys in WT table.
# - Relatively large number of read operation threads to stress cache.
# Benchmark is based on a wtperf config: wiredtiger/bench/wtperf/runners/evict-btree-scan.wtperf
# There are number of changes made to original configs such as:
# - Reduced the number of documents inserts during initial warm-up phase.
# - Increased the sizes of key and value.
# - Reduced the WT cache size.
# - Added transaction based operations and repurposed some of the read threads as long-running
# transcation threads.
###################################################################################################
'''
# wtperf options file: evict btree configuration
conn_config="cache_size=40G,checkpoint=(wait=60,log_size=2GB),eviction=(threads_min=12,
threads_max=12),log=(enabled=true),session_max=600,eviction_target=60,statistics=(fast),
statistics_log=(wait=1,json)"
# 1B records * (key=12 + value=138) is about 150G total data size
key_sz=12
value_sz=138
log_like_table=true
table_config="type=file"
icount=1000000000
report_interval=5
run_time=3600
# Scans every 10 minutes for all the scan specific tables.
# .4B records * (key=12 + value=138) is about 60G total data size for scan
# Running on a machine with 64G physical memory, this exhausts both the
# WT cache and the system cache.
scan_interval=600
scan_pct=100
scan_table_count=20
scan_icount=400000000
populate_threads=5
table_count=100
threads=((count=400,reads=1),(count=20,inserts=1,throttle=500),(count=10,updates=1,throttle=500))
# Add throughput/latency monitoring
max_latency=50000
sample_interval=5
'''
###################################################################################################
from runner import *
from wiredtiger import *
from workgen import *
context = Context()
homedir = "WT_TEST"
conn_config = "cache_size=1G,checkpoint=(wait=60,log_size=2GB),\
eviction=(threads_min=12,threads_max=12),log=(enabled=true),session_max=800,\
eviction_target=60,statistics=(fast),statistics_log=(wait=1,json)"# explicitly added
conn = wiredtiger_open(homedir, "create," + conn_config)
s = conn.open_session("")
wtperf_table_config = "key_format=S,value_format=S," +\
"exclusive=true,allocation_size=4kb," +\
"internal_page_max=64kb,leaf_page_max=4kb,split_pct=100,"
compress_table_config = ""
table_config = "type=file"
tables = []
table_count = 1
for i in range(0, table_count):
tname = "table:test" + str(i)
table = Table(tname)
s.create(tname, wtperf_table_config +\
compress_table_config + table_config)
table.options.key_size = 200
table.options.value_size = 5000
tables.append(table)
populate_threads = 40
icount = 500000
# If there are multiple tables to be filled during populate,
# the icount is split between them all.
pop_ops = Operation(Operation.OP_INSERT, tables[0])
pop_ops = op_multi_table(pop_ops, tables)
nops_per_thread = icount // (populate_threads * table_count)
pop_thread = Thread(pop_ops * nops_per_thread)
pop_workload = Workload(context, populate_threads * pop_thread)
pop_workload.run(conn)
# Log like file, requires that logging be enabled in the connection config.
log_name = "table:log"
s.create(log_name, wtperf_table_config + "key_format=S,value_format=S," +\
compress_table_config + table_config + ",log=(enabled=true)")
log_table = Table(log_name)
ops = Operation(Operation.OP_SEARCH, tables[0],Key(Key.KEYGEN_PARETO, 0, ParetoOptions(1)))
ops = op_multi_table(ops, tables, False)
ops = op_log_like(ops, log_table, 0)
thread0 = Thread(ops)
ops = Operation(Operation.OP_INSERT, tables[0])
ops = op_multi_table(ops, tables, False)
ops = op_log_like(ops, log_table, 0)
thread1 = Thread(ops)
# These operations include log_like operations, which will increase the number
# of insert/update operations by a factor of 2.0. This may cause the
# actual operations performed to be above the throttle.
ops = Operation(Operation.OP_UPDATE, tables[0])
ops = op_multi_table(ops, tables, False)
ops = op_log_like(ops, log_table, 0)
thread2 = Thread(ops)
# These operations include log_like operations, which will increase the number
# of insert/update operations by a factor of 2.0. This may cause the
# actual operations performed to be above the throttle.
thread2.options.throttle=500
thread2.options.throttle_burst=1.0
# Long running transactions. There is a 0.1 second sleep after a series of search and update
# operations. The sleep op is repeated 10000 times and this will make these transcations to at
# least run for ~17 minutes.
search_op = Operation(Operation.OP_SEARCH, tables[0], Key(Key.KEYGEN_PARETO, 0, ParetoOptions(1)))
update_op = Operation(Operation.OP_UPDATE, tables[0], Key(Key.KEYGEN_PARETO, 0, ParetoOptions(1)))
ops = txn(((search_op + update_op) * 1000 + sleep(0.1)) * 10000)
thread3 = Thread(ops)
ops = Operation(Operation.OP_SLEEP, "0.1") + \
Operation(Operation.OP_LOG_FLUSH, "")
logging_thread = Thread(ops)
workload = Workload(context, 400 * thread0 + 100 * thread1 +\
10 * thread2 + 100 * thread3 + logging_thread)
workload.options.report_interval=5
workload.options.run_time=500
workload.options.max_latency=50000
workload.run(conn)
latency_filename = homedir + "/latency.out"
latency.workload_latency(workload, latency_filename)
conn.close()
|
