3 RELEASE NOTES FOR TUXERA BENCHMARK
8 Tuxera.bench is a high-performance benchmark, similar to 'iozone'.
9 The major difference is that while 'iozone' can run for a long time
10 to collect performance results, tuxera.bench can get the same in a
13 Tuxera.bench by default runs a test to determine the cached and
14 effective read and write speeds using different I/O buffer sizes in a
15 very short time. The benchmark tool can be used to find the optimal
16 I/O buffer size and alignment giving the highest throughput on a
17 platform to tune performance critical applications.
19 With --speedtest option it runs a different test, which takes speed
20 samples at fixed intervals while writting a file to disk.
24 tuxera.bench [OPTIONS]
26 Benchmark streaming file I/O.
28 -a, --alignment NUMBER Buffer alignment [default: 4096]
29 -d, --directio No driver caching (O_DIRECT)
30 -b, --begin NUMBER Minimum block size [default: 512]
31 -e, --end NUMBER Maximum block size [default: 4 MB]
32 -t, --time NUMBER Time to run each test [default: 1 sec]
33 -s, --speedtest Run speed test
34 -B, --blocksize NUMBER Block size for speed test [default: 1048576 bytes]
35 -f, --filesize NUMBER File size for speed test [default: 1024 MB]
36 -h, --help Display this help
38 Typically the default buffer alignment gives the best performance
39 which is automatically determined.
41 Usage: 'cd' into a directory mounted by a file system driver to be
42 tested then run tuxera.bench.
44 During comparative testing make sure the tests are done using
45 the same disk and the same partition using different file systems.
46 Otherwise the comparison can be incorrect and misleading.
48 Flash memories (USB pen drives, SD, MMC, CF, etc) can have a
49 significantly lower speed limit than the maximum speed achievable
50 by the file system drivers and the results can vary in between
51 test runs. Using the '--time 10' tuxera.bench command line option
52 is recommended to get balanced and reproducible results.
54 Example output for default test:
56 *===================================================================*
57 | | I/O Block Performance |
58 | |-----------------------------------------------------------|
59 | Block | Write (MB/sec) | Read (MB/sec) |
60 | Size |-----------------------|-----------------------------------|
61 | | Cached | Effective | Cached | Effective | Device |
62 *===================================================================*
63 512 81.66 45.49 219.12 96.79 106.70
64 1024 112.90 74.70 298.66 97.06 107.49
65 2048 137.34 79.63 358.69 98.09 106.96
66 4096 148.21 83.78 380.33 97.78 103.79
67 8192 163.32 90.54 393.68 96.63 103.41
68 16384 162.68 89.51 400.86 96.38 106.03
69 32768 158.44 90.91 404.39 96.20 105.08
70 65536 163.84 90.12 408.33 95.87 104.47
71 131072 159.15 86.82 399.78 94.94 106.15
72 262144 164.15 88.45 415.99 97.37 106.72
73 524288 155.40 84.95 421.55 97.66 106.08
74 1048576 159.02 85.83 425.46 96.80 104.81
75 2097152 163.75 87.80 427.69 96.63 106.60
76 4194304 156.83 84.09 426.02 96.51 106.50
77 ---------------------------------------------------------------------
78 Average 149.05 83.04 384.32 96.76 105.77
79 Maximum 164.15 90.91 427.69 98.09 107.49
80 *===================================================================*
81 | I/O Metadata Performance |
82 *===================================================================*
84 cached io/sec: 2074972
86 Explanation of the outputs:
88 'Cached Write Performance' means the performance what applications
89 and users experience if lots of free RAM is available but there is
90 no guarantee that the data is on the disk.
92 'Effective Write Performance' means that all data is written to the
93 disk. This is more relevant for sustained write (large amount of data).
95 'Cached Read Performance' typically shows the memory bandwidth.
96 Caching read data can greatly help small reads and short-term rereads.
98 'Effective Read Performance' means non-cached read performance from the
99 disk via the file system.
101 'Device Read Performance' means non-cached read performance from the
102 disk without file system involvement.
105 Example output of speedtest (invoked with --speedtest):
107 Requested file size (bytes): 2147483648
108 File size (MB): 1 Speed (MB/s): 264.76
109 File size (MB): 2 Speed (MB/s): 263.96
110 File size (MB): 4 Speed (MB/s): 266.08
111 File size (MB): 8 Speed (MB/s): 261.78
112 File size (MB): 16 Speed (MB/s): 275.21
113 File size (MB): 32 Speed (MB/s): 270.33
114 File size (MB): 64 Speed (MB/s): 9.78
115 File size (MB): 128 Speed (MB/s): 6.66
116 File size (MB): 256 Speed (MB/s): 5.74
117 File size (MB): 512 Speed (MB/s): 5.43
118 File size (MB): 1024 Speed (MB/s): 5.26
120 Explanation of the outputs:
122 'File size' represents the sampling point at which speed is
125 'Speed' indicates the speed at which data is being written to disk.
127 --- Contact & Support Information ---
129 Please send feedback and support queries to support@tuxera.com.
133 Copyright (c) 2008-2013 Tuxera Inc. All Rights Reserved.