yuu_nkjm blog

2011-08-04(Thu) 編集

[KVM][openSUSE][Linux] KVMのパフォーマンスをunixbenchで計測する

unixbenchのインストール

yastでレポジトリをbenchで検索するとcpufreq-bench, unixbench, dbench, xbenchが引っかかったが，unixbenchが一番良さそうだった．yastでインストールして，/usr/share/qa/unixbench/Runを実行すれば，ベンチマークが動き出す．

openSUSE11.4では，レポジトリに入っているのはunixbench 4.0.1という古いバージョンだった．5.x系はマルチプロセッサに対応しているのに対して，4系ではマルチプロセッサに対応していないことに注意．

KVMのパフォーマンス - Unix Bench 5.x系で計測

まとめ

実機とKVMで大きな差が見られたのは，Process Creationか．Pipe-based Context Switchingは，差が見られるときと見られないときがあって，イマイチはっきりしないが，差がありそう．
KVM上で動かす実アプリケーションに依存するとは思うが，実機ではなくKVMを使うのもあり．

実機(Core i7 950 @ 3.07GHz)

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)
 
   System: redwhale: openSUSE 11.4 (x86_64)
   OS: GNU/Linux -- 2.6.37.6-0.7-desktop -- #1 SMP PREEMPT 2011-07-21 02:17:24 +0200
   Machine: x86_64 (x86_64)
   CPU 0: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz (6146.5 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext,
          SYSENTER/SYSEXIT, SYSCALL/SYSRET, Intel virtualization
 (snip)
   CPU 7: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz (6207.5 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext,
          SYSENTER/SYSEXIT, SYSCALL/SYSRET, Intel virtualization
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 08:24:14 - 08:52:25
8 CPUs in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables       31286938.9 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     4935.1 MWIPS (9.8 s, 7 samples)
Execl Throughput                               2173.1 lps   (29.8 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1164790.4 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          342708.4 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2506433.3 KBps  (30.0 s, 2 samples)
Pipe Throughput                             2602655.6 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 224412.5 lps   (10.0 s, 7 samples)
Process Creation                              15829.8 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   5455.3 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   3380.0 lpm   (60.0 s, 2 samples)
System Call Overhead                        4519616.0 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   31286938.9   2681.0
Double-Precision Whetstone                       55.0       4935.1    897.3
Execl Throughput                                 43.0       2173.1    505.4
File Copy 1024 bufsize 2000 maxblocks          3960.0    1164790.4   2941.4
File Copy 256 bufsize 500 maxblocks            1655.0     342708.4   2070.7
File Copy 4096 bufsize 8000 maxblocks          5800.0    2506433.3   4321.4
Pipe Throughput                               12440.0    2602655.6   2092.2
Pipe-based Context Switching                   4000.0     224412.5    561.0
Process Creation                                126.0      15829.8   1256.3
Shell Scripts (1 concurrent)                     42.4       5455.3   1286.6
Shell Scripts (8 concurrent)                      6.0       3380.0   5633.4
System Call Overhead                          15000.0    4519616.0   3013.1
                                                                   ========
System Benchmarks Index Score                                        1782.7
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 08:52:25 - 09:20:45
8 CPUs in system; running 8 parallel copies of tests
 
Dhrystone 2 using register variables      134976120.3 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                    34135.1 MWIPS (9.7 s, 7 samples)
Execl Throughput                              25045.8 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        936668.4 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          254989.5 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2579818.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                            10666961.4 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                1929033.6 lps   (10.0 s, 7 samples)
Process Creation                              63909.4 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                  31369.3 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   4530.2 lpm   (60.1 s, 2 samples)
System Call Overhead                       15812586.1 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0  134976120.3  11566.1
Double-Precision Whetstone                       55.0      34135.1   6206.4
Execl Throughput                                 43.0      25045.8   5824.6
File Copy 1024 bufsize 2000 maxblocks          3960.0     936668.4   2365.3
File Copy 256 bufsize 500 maxblocks            1655.0     254989.5   1540.7
File Copy 4096 bufsize 8000 maxblocks          5800.0    2579818.8   4448.0
Pipe Throughput                               12440.0   10666961.4   8574.7
Pipe-based Context Switching                   4000.0    1929033.6   4822.6
Process Creation                                126.0      63909.4   5072.2
Shell Scripts (1 concurrent)                     42.4      31369.3   7398.4
Shell Scripts (8 concurrent)                      6.0       4530.2   7550.3
System Call Overhead                          15000.0   15812586.1  10541.7
                                                                   ========
System Benchmarks Index Score                                        5540.7

KVM プロセッサ×2(on Core i7 950 @ 3.07GHz)

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)
 
   System: kvmsuse: openSUSE 11.4 (x86_64)
   OS: GNU/Linux -- 2.6.37.6-0.7-desktop -- #1 SMP PREEMPT 2011-07-21 02:17:24 +0200
   Machine: x86_64 (x86_64)
   CPU 0: QEMU Virtual CPU version 0.14.0 (6146.5 bogomips)
          x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   CPU 1: QEMU Virtual CPU version 0.14.0 (6146.5 bogomips)
          x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 10:37:50 - 11:06:02
2 CPUs in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables       34825586.4 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     4947.3 MWIPS (9.8 s, 7 samples)
Execl Throughput                               2858.2 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1103293.7 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          314596.2 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2311162.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                             2497174.9 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 136686.3 lps   (10.0 s, 7 samples)
Process Creation                               8086.2 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   6008.4 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1368.4 lpm   (60.1 s, 2 samples)
System Call Overhead                        4473280.8 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   34825586.4   2984.2
Double-Precision Whetstone                       55.0       4947.3    899.5
Execl Throughput                                 43.0       2858.2    664.7
File Copy 1024 bufsize 2000 maxblocks          3960.0    1103293.7   2786.1
File Copy 256 bufsize 500 maxblocks            1655.0     314596.2   1900.9
File Copy 4096 bufsize 8000 maxblocks          5800.0    2311162.8   3984.8
Pipe Throughput                               12440.0    2497174.9   2007.4
Pipe-based Context Switching                   4000.0     136686.3    341.7
Process Creation                                126.0       8086.2    641.8
Shell Scripts (1 concurrent)                     42.4       6008.4   1417.1
Shell Scripts (8 concurrent)                      6.0       1368.4   2280.6
System Call Overhead                          15000.0    4473280.8   2982.2
                                                                   ========
System Benchmarks Index Score                                        1526.2
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 11:06:02 - 11:34:12
2 CPUs in system; running 2 parallel copies of tests
 
Dhrystone 2 using register variables       66933853.9 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     9533.4 MWIPS (9.8 s, 7 samples)
Execl Throughput                               8193.7 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1317719.0 KBps  (30.1 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          351611.6 KBps  (30.1 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       3256364.4 KBps  (30.0 s, 2 samples)
Pipe Throughput                             4737484.1 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 708496.1 lps   (10.0 s, 7 samples)
Process Creation                              21420.8 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   9374.1 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   1427.7 lpm   (60.1 s, 2 samples)
System Call Overhead                        6949687.6 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   66933853.9   5735.5
Double-Precision Whetstone                       55.0       9533.4   1733.3
Execl Throughput                                 43.0       8193.7   1905.5
File Copy 1024 bufsize 2000 maxblocks          3960.0    1317719.0   3327.6
File Copy 256 bufsize 500 maxblocks            1655.0     351611.6   2124.5
File Copy 4096 bufsize 8000 maxblocks          5800.0    3256364.4   5614.4
Pipe Throughput                               12440.0    4737484.1   3808.3
Pipe-based Context Switching                   4000.0     708496.1   1771.2
Process Creation                                126.0      21420.8   1700.1
Shell Scripts (1 concurrent)                     42.4       9374.1   2210.9
Shell Scripts (8 concurrent)                      6.0       1427.7   2379.5
System Call Overhead                          15000.0    6949687.6   4633.1
                                                                   ========
System Benchmarks Index Score                                        2776.8

KVM プロセッサ×6(on Core i7 950 @ 3.07GHz)

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)
 
   System: kvmsuse: openSUSE 11.4 (x86_64)
   OS: GNU/Linux -- 2.6.37.6-0.7-desktop -- #1 SMP PREEMPT 2011-07-21 02:17:24 +0200
   Machine: x86_64 (x86_64)
   CPU 0: QEMU Virtual CPU version 0.14.0 (6146.5 bogomips)
          x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
 (snip)
   CPU 5: QEMU Virtual CPU version 0.14.0 (6146.5 bogomips)
          x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
 
------------------------------------------------------------------------
Benchmark Run: 木  8月 18 2011 21:52:39 - 22:21:20
6 CPUs in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables       34727302.1 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     4948.3 MWIPS (9.8 s, 7 samples)
Execl Throughput                               1669.8 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1067639.6 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          297204.9 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2268963.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                             2598809.9 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 252469.2 lps   (10.0 s, 7 samples)
Process Creation                               3383.3 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   4625.3 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   2351.5 lpm   (60.1 s, 2 samples)
System Call Overhead                        4471679.3 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   34727302.1   2975.8
Double-Precision Whetstone                       55.0       4948.3    899.7
Execl Throughput                                 43.0       1669.8    388.3
File Copy 1024 bufsize 2000 maxblocks          3960.0    1067639.6   2696.1
File Copy 256 bufsize 500 maxblocks            1655.0     297204.9   1795.8
File Copy 4096 bufsize 8000 maxblocks          5800.0    2268963.8   3912.0
Pipe Throughput                               12440.0    2598809.9   2089.1
Pipe-based Context Switching                   4000.0     252469.2    631.2
Process Creation                                126.0       3383.3    268.5
Shell Scripts (1 concurrent)                     42.4       4625.3   1090.9
Shell Scripts (8 concurrent)                      6.0       2351.5   3919.1
System Call Overhead                          15000.0    4471679.3   2981.1
                                                                   ========
System Benchmarks Index Score                                        1453.4
 
------------------------------------------------------------------------
Benchmark Run: 木  8月 18 2011 22:21:20 - 22:49:32
6 CPUs in system; running 6 parallel copies of tests
 
Dhrystone 2 using register variables      132040998.4 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                    26264.2 MWIPS (9.7 s, 7 samples)
Execl Throughput                              16806.4 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        671883.7 KBps  (30.1 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          176490.2 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2219914.8 KBps  (30.0 s, 2 samples)
Pipe Throughput                             8925354.7 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                1266987.8 lps   (10.0 s, 7 samples)
Process Creation                              37824.7 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                  21340.7 lpm   (60.1 s, 2 samples)
Shell Scripts (8 concurrent)                   3018.2 lpm   (60.2 s, 2 samples)
System Call Overhead                       14155711.6 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0  132040998.4  11314.6
Double-Precision Whetstone                       55.0      26264.2   4775.3
Execl Throughput                                 43.0      16806.4   3908.5
File Copy 1024 bufsize 2000 maxblocks          3960.0     671883.7   1696.7
File Copy 256 bufsize 500 maxblocks            1655.0     176490.2   1066.4
File Copy 4096 bufsize 8000 maxblocks          5800.0    2219914.8   3827.4
Pipe Throughput                               12440.0    8925354.7   7174.7
Pipe-based Context Switching                   4000.0    1266987.8   3167.5
Process Creation                                126.0      37824.7   3002.0
Shell Scripts (1 concurrent)                     42.4      21340.7   5033.2
Shell Scripts (8 concurrent)                      6.0       3018.2   5030.3
System Call Overhead                          15000.0   14155711.6   9437.1
                                                                   ========
System Benchmarks Index Score                                        4118.8

実機(QuadCore Xeon X3220 2.40GHz)

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.3)
 
   System: GNU/Linux
   OS: GNU/Linux -- 2.6.23.1-42.fc8 -- #1 SMP Tue Oct 30 13:55:12 EDT 2007
   Machine: i686 (i386)
   CPU 0: Intel(R) Xeon(R) CPU X3220 @ 2.40GHz (4802.3 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, Intel virtualization
 (snip)
   CPU 3: Intel(R) Xeon(R) CPU X3220 @ 2.40GHz (4799.5 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, Intel virtualization
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 08:23:01 - 08:51:06
4 CPUs in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables       10538103.7 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     2204.4 MWIPS (10.2 s, 7 samples)
Execl Throughput                               2390.1 lps   (29.8 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        344724.4 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           88277.9 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        825631.4 KBps  (30.0 s, 2 samples)
Pipe Throughput                              492919.4 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                  80393.3 lps   (10.0 s, 7 samples)
Process Creation                               7785.5 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   4017.6 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   2163.5 lpm   (60.0 s, 2 samples)
System Call Overhead                         648029.3 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   10538103.7    903.0
Double-Precision Whetstone                       55.0       2204.4    400.8
Execl Throughput                                 43.0       2390.1    555.8
File Copy 1024 bufsize 2000 maxblocks          3960.0     344724.4    870.5
File Copy 256 bufsize 500 maxblocks            1655.0      88277.9    533.4
File Copy 4096 bufsize 8000 maxblocks          5800.0     825631.4   1423.5
Pipe Throughput                               12440.0     492919.4    396.2
Pipe-based Context Switching                   4000.0      80393.3    201.0
Process Creation                                126.0       7785.5    617.9
Shell Scripts (1 concurrent)                     42.4       4017.6    947.6
Shell Scripts (8 concurrent)                      6.0       2163.5   3605.8
System Call Overhead                          15000.0     648029.3    432.0
                                                                   ========
System Benchmarks Index Score                                         679.3
 
------------------------------------------------------------------------
Benchmark Run: 金  8月 19 2011 08:51:06 - 09:19:12
4 CPUs in system; running 4 parallel copies of tests
 
Dhrystone 2 using register variables       41129963.1 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     8437.9 MWIPS (10.1 s, 7 samples)
Execl Throughput                              12350.5 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        145095.5 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           39586.0 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        484689.7 KBps  (30.0 s, 2 samples)
Pipe Throughput                             1949629.1 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 678733.6 lps   (10.0 s, 7 samples)
Process Creation                              37720.0 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                  18309.8 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                   2512.9 lpm   (60.0 s, 2 samples)
System Call Overhead                        2049215.9 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   41129963.1   3524.4
Double-Precision Whetstone                       55.0       8437.9   1534.2
Execl Throughput                                 43.0      12350.5   2872.2
File Copy 1024 bufsize 2000 maxblocks          3960.0     145095.5    366.4
File Copy 256 bufsize 500 maxblocks            1655.0      39586.0    239.2
File Copy 4096 bufsize 8000 maxblocks          5800.0     484689.7    835.7
Pipe Throughput                               12440.0    1949629.1   1567.2
Pipe-based Context Switching                   4000.0     678733.6   1696.8
Process Creation                                126.0      37720.0   2993.7
Shell Scripts (1 concurrent)                     42.4      18309.8   4318.3
Shell Scripts (8 concurrent)                      6.0       2512.9   4188.2
System Call Overhead                          15000.0    2049215.9   1366.1
                                                                   ========
System Benchmarks Index Score                                        1569.7

KVMのパフォーマンス - Unix Bench 4.x系で計測

4.x系の結果を残してあるが，5.x系を使った方が良い．

実機とKVMで大きな差が出たのは，Process Creationか．プロセッサを2個割り当てたときと比べて，プロセッサを6個割り当てたとき，Process Creationの性能が劣化したのは何でだろうか．あと，Pipe-based Context Switchingにも若干の性能劣化があるように見える．気のせいかもしれんけど．

実機(Core i7 950 @ 3.07GHz)

実機のデータは以下．プロセッサはCorei7 940(8コア)である．

BYTE UNIX Benchmarks (Version 4.0.1)
 
System -- Linux redwhale 2.6.37.6-0.7-desktop #1 SMP PREEMPT 2011-07-21 02:17:24
+0200 x86_64 x86_64 x86_64 GNU/Linux
 
TEST                                        BASELINE     RESULT      INDEX
Arithmetic Test (type = double)              29820.0        1.0        0.0
Dhrystone 2 using register variables        116700.0 22917423.2     1963.8
Execl Throughput                                43.0        0.0        0.0
File Copy 1024 bufsize 2000 maxblocks         3960.0  1268131.0     3202.4
File Copy 256 bufsize 500 maxblocks           1655.0   354522.0     2142.1
File Copy 4096 bufsize 8000 maxblocks         5800.0  2599929.0     4482.6
Pipe Throughput                              12440.0  2555097.3     2053.9
Pipe-based Context Switching                  4000.0   215490.4      538.7
Process Creation                               126.0    15677.9     1244.3
Shell Scripts (8 concurrent)                     6.0     2957.3     4928.8
System Call Overhead                         15000.0  4513453.5     3009.0

KVM プロセッサ×2(on Core i7 950 @ 3.07GHz)

KVMにプロセッサを2個割り当てた場合のデータは以下．

BYTE UNIX Benchmarks (Version 4.0.1)
 
System -- Linux kvmsuse 2.6.37.1-1.2-desktop #1 SMP PREEMPT 2011-02-21 10:34:10
+0100 x86_64 x86_64 x86_64 GNU/Linux
 
TEST                                        BASELINE     RESULT      INDEX
Arithmetic Test (type = double)              29820.0        1.0        0.0
Dhrystone 2 using register variables        116700.0 23293754.4     1996.0
Execl Throughput                                43.0        0.0        0.0
File Copy 1024 bufsize 2000 maxblocks         3960.0  1145196.0     2891.9
File Copy 256 bufsize 500 maxblocks           1655.0   321544.0     1942.9
File Copy 4096 bufsize 8000 maxblocks         5800.0  2510289.0     4328.1
Pipe Throughput                              12440.0  2525370.0     2030.0
Pipe-based Context Switching                  4000.0   125306.9      313.3
Process Creation                               126.0     7853.1      623.3
Shell Scripts (8 concurrent)                     6.0     1209.3     2015.5
System Call Overhead                         15000.0  4491089.4     2994.1

KVM プロセッサ×6(on Core i7 950 @ 3.07GHz)

KVMにプロセッサを6個割り当てた場合のデータは以下．

BYTE UNIX Benchmarks (Version 4.0.1)
System -- Linux kvmsuse 2.6.37.6-0.7-desktop #1 SMP PREEMPT 2011-07-21 02:17:24 +0200
x86_64 x86_64 x86_64 GNU/Linux
 
TEST                                        BASELINE     RESULT      INDEX
Arithmetic Test (type = double)              29820.0        1.0        0.0
Dhrystone 2 using register variables        116700.0 23093871.0     1978.9
Execl Throughput                                43.0        0.0        0.0
File Copy 1024 bufsize 2000 maxblocks         3960.0  1090723.0     2754.4
File Copy 256 bufsize 500 maxblocks           1655.0   297632.0     1798.4
File Copy 4096 bufsize 8000 maxblocks         5800.0  2294295.0     3955.7
Pipe Throughput                              12440.0  2500653.4     2010.2
Pipe-based Context Switching                  4000.0   268735.2      671.8
Process Creation                               126.0     3427.6      272.0
Shell Scripts (8 concurrent)                     6.0     2128.7     3547.8
System Call Overhead                         15000.0  4471322.5     2980.9

実機(QuadCore Xeon X3220 2.40GHz)

BYTE UNIX Benchmarks (Version 4.0.1)
System -- Linux falcon 2.6.23.1-42.fc8 #1 SMP Tue Oct 30 13:55:12 EDT 2007
i686 i686 i386 GNU/Linux
 
TEST                                        BASELINE     RESULT      INDEX
Arithmetic Test (type = double)              29820.0   642874.8      215.6
Dhrystone 2 using register variables        116700.0  9713051.3      832.3
Execl Throughput                                43.0     2387.3      555.2
File Copy 1024 bufsize 2000 maxblocks         3960.0   304295.0      768.4
File Copy 256 bufsize 500 maxblocks           1655.0    86306.0      521.5
File Copy 4096 bufsize 8000 maxblocks         5800.0   806413.0     1390.4
Pipe Throughput                              12440.0   477059.3      383.5
Pipe-based Context Switching                  4000.0    88368.5      220.9
Process Creation                               126.0     7726.8      613.2
Shell Scripts (8 concurrent)                     6.0     2272.0     3786.7
System Call Overhead                         15000.0   645915.3      430.6

[ツッコミを入れる]

2012-08-04(Sat) 編集

[Network][arp][route] BuffaloのブリッジモードのBBルータに無理矢理接続する

注意

書きかけ．下記の手順でBBルータへパケットが届くようになるとは思うが，帰りがどうなっているのか不明．BBルータをブリッヂモードにしているから帰ってくるのかなー．

概要

同一ネットワーク内に存在する異なるネットワークアドレスを持つbuffaloのBBルータに接続する．

ホストPCのIPアドレスは10.100.100.213，BBルータはデフォルトの設定で192.168.11.100をIPアドレスとして持っている．ホストPCとBBルータが別のネットワークにあるため，Web設定画面に接続出来ない．

補足事項として，ホストPCは無線LANと有線LANが使え，ホストPCは有線LANで繋いであるものとする．

まず，arpテーブルを見るとこんな感じ．

$ arp -a 
  10.100.100.213        xx-xx-xx-xx-xx-xx     動的
  10.100.100.254        xx-xx-xx-xx-xx-xx     動的
  10.100.100.255        xx-xx-xx-xx-xx-xx     静的
  192.168.11.100        00-xx-xx-xx-xx-51     動的

まず，ホストPCのarpテーブルを書きかえて，10.100.100.215と00-xx-xx-xx-xx-51を紐づける作戦を考えた．．ホストPCから10.100.100.215へアクセスしようとすると，ネットワークアドレスのチェックより先にarpテーブルが検索され，00-xx-xx-xx-xx-51と通信が始まることを期待．

$ arp -s 10.100.100.215 00-xx-xx-xx-xx-51
# arpテーブルを書きかえる．
 
$ arp -a
  インターフェイス: 10.100.100.213 --- 0xc
  インターネット アドレス      物理アドレス      種類
  10.100.100.213        xx-xx-xx-xx-xx-xx     動的
  10.100.100.215        00-xx-xx-xx-xx-51     静的
  10.100.100.254        xx-xx-xx-xx-xx-xx     動的
  10.100.100.255        xx-xx-xx-xx-xx-xx     静的
  192.168.11.100        00-xx-xx-xx-xx-51     動的
# 書き換わってる．

arpテーブルを書きかえた後，ホストPCから10.100.100.215へ接続したが，BBルータのWeb設定画面に接続出来なかった．(インタフェースの設定をしていなかったのが関係している？)

arpテーブルを書きかえるのではなく，ルーティングテーブルを書きかえてみる．

書きかえ前．

route print
 
IPv4 ルート テーブル
===========================================================================
アクティブ ルート:
ネットワーク宛先   ネットマスク       ゲートウェイ     インターフェイス  メトリック
0.0.0.0            0.0.0.0            10.100.100.254   10.100.100.213    266
10.100.100.0       255.255.255.0      リンク上         10.100.100.213    266
10.100.100.213     255.255.255.255    リンク上         10.100.100.213    266
10.100.100.255     255.255.255.255    リンク上         10.100.100.213    266
===========================================================================

192.168.11.100というネットワーク(1台だけのネットワークなので255.255.255.255)へのアクセスを同一ネットワークへのアクセスと同じようにさせる．どうしたら良いのか分からないけど，ひとまずデフォルトゲートウェイにいってしまわないようにしてみる．

route add 192.168.11.100 mask 255.255.255.255 0.0.0.0 metric 1 if 12
# 12は"route print"で上の方に表示されたインタフェース番号．
 
route print
 
IPv4 ルート テーブル
===========================================================================
アクティブ ルート:
ネットワーク宛先  ネットマスク     ゲートウェイ      インターフェイス     メトリック
0.0.0.0           0.0.0.0          10.100.100.254    10.100.100.213    266
10.100.100.0      255.255.255.0       リンク上       10.100.100.213    266
10.100.100.213    255.255.255.255     リンク上       10.100.100.213    266
10.100.100.255    255.255.255.255     リンク上       10.100.100.213    266
192.168.11.100    255.255.255.255  0.0.0.0           10.100.100.213    1
===========================================================================
#リモートセッションが切れてメモできなかったので，記憶から再現．

routeテーブルを書きかえると，10.100.100.215から192.168.11.100にアクセスできるようになった．