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
|
# Usage
This benchmark is a server-side SSLSocket benchmark for use with clients
which measure performance. This benchmark does not, itself, make any
performance measurements. For instance:
```bash
$ ./run_test.sh org.mozilla.jss.tests.BenchmarkSSLSocket JSS.legacy Server_RSA 8443 1024 &
$ siege -c 100 -b -t 5m https://localhost:8443
$ kill %1
```
There are three supported SSLSocket implementations:
1. `org.mozilla.jss.ssl.SSLSocket`, JSS's legacy implementation
name: `JSS.legacy`
2. `org.mozilla.jss.ssl.javax.JSSSocket`, JSS's new javax implementation
name: `JSS.SSLSocket`
3. `sun.security`'s `SSLSocketImpl` from the current JDK.
name: `SunJSSE.SSLSocket`
It is suggested to disable all logging (for instance, via:
`truncate -s 0 tools/logging.properties`) in order have reproducible
results.
This class takes four arguments when invoked:
1. The name of the implementation to benchmark, see above.
2. An alias of the certificate or path to a PKCS#12 file. Only SunJSSE
accepts a PKCS#12 as path -- the two JSS based SSLSocket
implementations will utilize a nickname instead.
3. The port to listen on.
4. The size of the HTTP message to fake.
Note that, when utilizing a JSS provider, JSS must be loaded via a
java.security. When utilizing SunJSSE, for best results, do not load
JSS via java.security.
It is suggested to use `run_test.sh` from the `build/` directory for
executing this utility.
# Past Performance
## `JSSEngineReferenceImpl`
These are the results from siege [0] as run via:
```bash
$ siege -c 100 -b -t 5m https://localhost:8443
```
with the benchmarker set to send a faked 1024-byte message:
```bash
$ ./run_test.sh org.mozilla.jss.tests.BenchmarkSSLSocket JSS.legacy Server_RSA 8443 1024
```
The server certificate is 4096-bits. The selection of cipher suite and
protocol is left at their defaults. This is on a Lenovo Thinkpad P50
with an `Intel(R) Core(TM) i7-6820HQ CPU @ 2.70GHz` processor and 32GB of RAM.
Each request gets spun off and handled by a new thread.
Using the legacy `org.mozilla.jss.ssl.SSLSocket` (old NSS-based socket)
memory stays stable and under ~1-2% of total memory:
```json
{
"transactions": 98588,
"availability": 100.00,
"elapsed_time": 299.43,
"data_transferred": 96.28,
"response_time": 0.30,
"transaction_rate": 329.25,
"throughput": 0.32,
"concurrency": 99.80,
"successful_transactions": 98588,
"failed_transactions": 0,
"longest_transaction": 29.94,
"shortest_transaction": 0.04
}
```
Using `javax.net.ssl.SSLSocket` provided by SunJSSE (but with JSS crypto
and potentially random) and stays under 1-2% of total memory:
```json
{
"transactions": 2417,
"availability": 100.00,
"elapsed_time": 299.36,
"data_transferred": 2.36,
"response_time": 12.12,
"transaction_rate": 8.07,
"throughput": 0.01,
"concurrency": 97.82,
"successful_transactions": 2417,
"failed_transactions": 0,
"longest_transaction": 21.27,
"shortest_transaction": 1.63
}
```
**Note** that the above option was removed from the benchmark utility as it
was significantly slower.
Using `javax.net.ssl.SSLSocket` provided by SunJSSE (without JSS crypto,
via exporting to PKCS12 file) and stays under 8% of total memory:
```json
{
"transactions": 93168,
"availability": 100.00,
"elapsed_time": 299.92,
"data_transferred": 90.98,
"response_time": 0.32,
"transaction_rate": 310.64,
"throughput": 0.30,
"concurrency": 99.51,
"successful_transactions": 93168,
"failed_transactions": 2,
"longest_transaction": 15.81,
"shortest_transaction": 0.02
}
```
And `javax.net.ssl.SSLSocket` provided by Mozilla-JSS, backed by our slow
JSSEngine (proposed for 8.3) -- memory grows to ~35% of total, which suggests
there's also at least one memory leak still...
```json
{
"transactions": 87768,
"availability": 100.00,
"elapsed_time": 299.08,
"data_transferred": 85.71,
"response_time": 0.34,
"transaction_rate": 293.46,
"throughput": 0.29,
"concurrency": 99.60,
"successful_transactions": 87768,
"failed_transactions": 1,
"longest_transaction": 16.05,
"shortest_transaction": 0.08
}
```
Prior to jss-pr#553 (commit 1bd646a45613d16f18f28c641381f680ba1df319), the
performance of Mozilla-JSS's `SSLSocket` was similar to
`javax.net.ssl.SSLSoket` using `Mozilla-JSS` for primitives:
```json
{
"transactions": 1551,
"availability": 85.98,
"elapsed_time": 299.53,
"data_transferred": 1.51,
"response_time": 13.02,
"transaction_rate": 5.18,
"throughput": 0.01,
"concurrency": 67.42,
"successful_transactions": 1551,
"failed_transactions": 253,
"longest_transaction": 78.01,
"shortest_transaction": 0.50
}
```
And for comparison, `nginx-1.18.0-1.fc32.x86_64`, using the same cert from
above (admittedly, it uses OpenSSL and an `epoll` framework) and same `siege`
output:
```json
{
"transactions": 214725,
"availability": 100.00,
"elapsed_time": 299.05,
"data_transferred": 209.90,
"response_time": 0.14,
"transaction_rate": 718.02,
"throughput": 0.70,
"concurrency": 99.40,
"successful_transactions": 214725,
"failed_transactions": 0,
"longest_transaction": 0.37,
"shortest_transaction": 0.07
}
```
|
