今晚试了试UnixBench在VPS上的测试，想比较下BurstNET和Linode的VPS的差异。首先要准备一下测试环境，下载和编译一下UnixBench。这次测试我分别用了UnixBench 4.1.0版，和最新的UnixBench 5.1.2版。下载及编译UnixBench:

wget http://members.dslextreme.com/users/andylee/unixbench-4.1.0-wht.tar.gz
tar zxvf unixbench-4.1.0-wht.tar.gz
#解压后，得到一个unixbench-4.1.0-wht-2的目录，进入目录，然后编译unixbench
cd unixbench-4.1.0-wht-2
make
#编译完成后，运行
./Run

这里还要注意一下，如果运行时出错，提示：Please install /usr/bin/time. 的话，需要安装time

apt-get install time

UnixBench 5.1.2版的编译方法类似，如果在编译时出现错误，提示找不到X11一类的库的时候，需要编辑Makefile，将 GRAPHIC_TESTS = defined 这一行注释掉即可，因为对于VPS，不需要对显卡进行测试。

这样，编译完成之后，就可以开始运行测试了。整个过程大概持续十几分钟，最终的结果就会出来。

BurstNET VPS 配置： CPU 单核 Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz 2MB二级缓存，512MB内存，20G硬盘 Linode VPS 配置： CPU 四核 Intel(R) Xeon(R) CPU L5520 @ 2.27GHz 每核8MB二级缓存，360MB内存，16 G硬盘

linode UnixBench4.1.0得分:

Start Benchmark Run: Sun Jun 13 23:52:56 CST 2010

23:52:56 up 7 days, 4:34, 2 users, load average: 0.84, 3.10, 2.03

End Benchmark Run: Mon Jun 14 00:03:44 CST 2010

00:03:44 up 7 days, 4:45, 2 users, load average: 15.07, 6.48, 3.71

INDEX VALUES TEST BASELINE RESULT INDEX

Dhrystone 2 using register variables 376783.7 18549224.1 492.3

Double-Precision Whetstone 83.1 1075.3 129.4

Execl Throughput 188.3 271.3 14.4

File Copy 1024 bufsize 2000 maxblocks 2672.0 183122.0 685.3

File Copy 256 bufsize 500 maxblocks 1077.0 46095.0 428.0

File Read 4096 bufsize 8000 maxblocks 15382.0 2248347.0 1461.7

Pipe Throughput 111814.6 1426039.0 127.5

Pipe-based Context Switching 15448.6 153447.5 99.3

Process Creation 569.3 4589.7 80.6

Shell Scripts (8 concurrent) 44.8 1257.8 280.8

System Call Overhead 114433.5 1138915.6 99.5

=========

FINAL SCORE 189.4

BurstNET UnixBench4.1.0得分：

Start Benchmark Run: Sun Jun 13 15:44:53 UTC 2010

15:44:53 up 17 days, 20:22, 1 user, load average: 0.61, 0.68, 0.43

End Benchmark Run: Sun Jun 13 16:11:34 UTC 2010

16:11:34 up 17 days, 20:49, 1 user, load average: 14.26, 6.00, 3.34

INDEX VALUES TEST BASELINE RESULT INDEX

Dhrystone 2 using register variables 376783.7 5393975.4 143.2

Double-Precision Whetstone 83.1 1229.0 147.9

Execl Throughput 188.3 3386.5 179.8

File Copy 1024 bufsize 2000 maxblocks 2672.0 172137.0 644.2

File Copy 256 bufsize 500 maxblocks 1077.0 67889.0 630.4

File Read 4096 bufsize 8000 maxblocks 15382.0 833087.0 541.6

Pipe-based Context Switching 15448.6 233051.2 150.9

Pipe Throughput 111814.6 650442.2 58.2

Process Creation 569.3 9861.9 173.2

Shell Scripts (8 concurrent) 44.8 317.0 70.8

System Call Overhead 114433.5 586353.5 51.2

=========

FINAL SCORE 174.7 Linode UnixBench 5.1.2得分：

BYTE UNIX Benchmarks (Version 5.1.2)

System: li165-157: GNU/Linux

OS: GNU/Linux – 2.6.32.12-linode25 – #1 SMP Wed Apr 28 19:25:11 UTC 2010

Machine: i686 (unknown)

Language: en_US.utf8 (charmap=”UTF-8”, collate=”UTF-8”)

CPU 0: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz (4533.5 bogomips)

Hyper-Threading, MMX, Physical Address Ext

CPU 1: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz (4533.5 bogomips)

Hyper-Threading, MMX, Physical Address Ext

CPU 2: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz (4533.5 bogomips)

Hyper-Threading, MMX, Physical Address Ext

CPU 3: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz (4533.5 bogomips)

Hyper-Threading, MMX, Physical Address Ext

00:10:48 up 7 days, 4:52, 2 users, load average: 0.08, 1.57, 2.34; runlevel 2

Benchmark Run: Mon Jun 14 2010 00:10:48 - 00:38:47

4 CPUs in system; running 1 parallel copy of tests

Dhrystone 2 using register variables 14795997.6 lps (10.0 s, 7 samples)

Double-Precision Whetstone 2104.9 MWIPS (10.2 s, 7 samples)

Execl Throughput 1114.3 lps (30.0 s, 2 samples)

File Copy 1024 bufsize 2000 maxblocks 289394.9 KBps (30.0 s, 2 samples)

File Copy 256 bufsize 500 maxblocks 76973.1 KBps (30.0 s, 2 samples)

File Copy 4096 bufsize 8000 maxblocks 807122.7 KBps (30.0 s, 2 samples)

Pipe Throughput 418521.2 lps (10.0 s, 7 samples)

Pipe-based Context Switching 21882.3 lps (10.0 s, 7 samples)

Process Creation 2082.0 lps (30.0 s, 2 samples)

Shell Scripts (1 concurrent) 2704.7 lpm (60.0 s, 2 samples)

Shell Scripts (8 concurrent) 703.4 lpm (60.0 s, 2 samples)

System Call Overhead 430386.2 lps (10.0 s, 7 samples)

System Benchmarks Index Values BASELINE RESULT INDEX

Dhrystone 2 using register variables 116700.0 14795997.6 1267.9

Double-Precision Whetstone 55.0 2104.9 382.7

Execl Throughput 43.0 1114.3 259.1

File Copy 1024 bufsize 2000 maxblocks 3960.0 289394.9 730.8

File Copy 256 bufsize 500 maxblocks 1655.0 76973.1 465.1

File Copy 4096 bufsize 8000 maxblocks 5800.0 807122.7 1391.6

Pipe Throughput 12440.0 418521.2 336.4

Pipe-based Context Switching 4000.0 21882.3 54.7

Process Creation 126.0 2082.0 165.2

Shell Scripts (1 concurrent) 42.4 2704.7 637.9

Shell Scripts (8 concurrent) 6.0 703.4 1172.4

System Call Overhead 15000.0 430386.2 286.9

========

System Benchmarks Index Score 429.0

Benchmark Run: Mon Jun 14 2010 00:38:47 - 01:07:04

4 CPUs in system; running 4 parallel copies of tests

Dhrystone 2 using register variables 51437080.2 lps (10.0 s, 7 samples)

Double-Precision Whetstone 7900.6 MWIPS (10.2 s, 7 samples)

Execl Throughput 4048.4 lps (29.9 s, 2 samples)

File Copy 1024 bufsize 2000 maxblocks 279522.6 KBps (30.0 s, 2 samples)

File Copy 256 bufsize 500 maxblocks 74109.8 KBps (30.0 s, 2 samples)

File Copy 4096 bufsize 8000 maxblocks 1020171.1 KBps (30.0 s, 2 samples)

Pipe Throughput 1495820.1 lps (10.0 s, 7 samples)

Pipe-based Context Switching 165631.7 lps (10.0 s, 7 samples)

Process Creation 4516.0 lps (30.0 s, 2 samples)

Shell Scripts (1 concurrent) 5624.7 lpm (60.0 s, 2 samples)

Shell Scripts (8 concurrent) 862.6 lpm (60.2 s, 2 samples)

System Call Overhead 1451308.5 lps (10.1 s, 7 samples)

System Benchmarks Index Values BASELINE RESULT INDEX

Dhrystone 2 using register variables 116700.0 51437080.2 4407.6

Double-Precision Whetstone 55.0 7900.6 1436.5

Execl Throughput 43.0 4048.4 941.5

File Copy 1024 bufsize 2000 maxblocks 3960.0 279522.6 705.9

File Copy 256 bufsize 500 maxblocks 1655.0 74109.8 447.8

File Copy 4096 bufsize 8000 maxblocks 5800.0 1020171.1 1758.9

Pipe Throughput 12440.0 1495820.1 1202.4

Pipe-based Context Switching 4000.0 165631.7 414.1

Process Creation 126.0 4516.0 358.4

Shell Scripts (1 concurrent) 42.4 5624.7 1326.6

Shell Scripts (8 concurrent) 6.0 862.6 1437.6

System Call Overhead 15000.0 1451308.5 967.5

========

System Benchmarks Index Score 1007.3

BurstNET UnixBench 5.1.2得分：

BYTE UNIX Benchmarks (Version 5.1.2)

System: vps.xiaozhou.net: GNU/Linux OS: GNU/Linux – 2.6.18-128.2.1.el5.028stab064.4 – #1 SMP Mon Jul 27 12:45:01 MSD 2009 Machine: i686 (unknown) Language: en_US.utf8 (charmap=”UTF-8”, collate=”UTF-8”) CPU 0: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz (4670.1 bogomips) Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET 02:29:02 up 18 days, 7:06, 1 user, load average: 0.16, 0.16, 0.17; runlevel 2

Benchmark Run: Mon Jun 14 2010 02:29:02 - 03:13:12 1 CPU in system; running 1 parallel copy of tests

Dhrystone 2 using register variables 11339207.7 lps (10.0 s, 7 samples) Double-Precision Whetstone 2260.5 MWIPS (10.2 s, 7 samples) Execl Throughput 3362.8 lps (29.6 s, 2 samples) File Copy 1024 bufsize 2000 maxblocks 319380.6 KBps (30.0 s, 2 samples) File Copy 256 bufsize 500 maxblocks 113284.5 KBps (30.0 s, 2 samples) File Copy 4096 bufsize 8000 maxblocks 807667.8 KBps (30.0 s, 2 samples) Pipe Throughput 684264.4 lps (10.0 s, 7 samples) Pipe-based Context Switching 234582.0 lps (10.0 s, 7 samples) Process Creation 9991.1 lps (30.1 s, 2 samples) Shell Scripts (1 concurrent) 2967.0 lpm (60.0 s, 2 samples) Shell Scripts (8 concurrent) 381.1 lpm (60.1 s, 2 samples) System Call Overhead 602393.1 lps (10.0 s, 7 samples)

System Benchmarks Index Values BASELINE RESULT INDEX Dhrystone 2 using register variables 116700.0 11339207.7 971.7 Double-Precision Whetstone 55.0 2260.5 411.0 Execl Throughput 43.0 3362.8 782.0 File Copy 1024 bufsize 2000 maxblocks 3960.0 319380.6 806.5 File Copy 256 bufsize 500 maxblocks 1655.0 113284.5 684.5 File Copy 4096 bufsize 8000 maxblocks 5800.0 807667.8 1392.5 Pipe Throughput 12440.0 684264.4 550.1 Pipe-based Context Switching 4000.0 234582.0 586.5 Process Creation 126.0 9991.1 792.9 Shell Scripts (1 concurrent) 42.4 2967.0 699.8 Shell Scripts (8 concurrent) 6.0 381.1 635.2

System Call Overhead 15000.0 602393.1 401.6

System Benchmarks Index Score 686.5

看来在单核CPU测试下，Linode并不占据太大优势，甚至得分低于BurstNET，估计是因为Linode内存比BurstNET小的缘故。在并行测试中，四核CPU明显有一定的优势，在能充分利用多核性能，所以得分较高。

最后，附上UnixBench在Google Code的地址： http://code.google.com/p/byte-unixbench/

后记

此文写于较早时候，当时Linode最低Plan的内存才300多MB，不过经过很多轮的升级，现在Linode已经完胜BurstNET了！