1The following are additional notes on the dtruss program, 2 3 4* Can I trust the elapsed and on-cpu times? 5 6Firstly, lets see dtruss do something cool, 7 8 # dtruss -eo -n bash 9 PID/LWP ELAPSD CPU SYSCALL(args) = return 10 6215/1: 57 37 write(0x2, "h\0", 0x1) = 1 0 11 6215/1: 357210 45 read(0x0, "e\0", 0x1) = 1 0 12 6215/1: 53 37 write(0x2, "e\0", 0x1) = 1 0 13 6215/1: 359510 46 read(0x0, "l\0", 0x1) = 1 0 14 6215/1: 57 42 write(0x2, "l\0", 0x1) = 1 0 15 6215/1: 166495 47 read(0x0, "l\0", 0x1) = 1 0 16 6215/1: 56 40 write(0x2, "l\0", 0x1) = 1 0 17 6215/1: 346076 44 read(0x0, "o\0", 0x1) = 1 0 18 6215/1: 54 38 write(0x2, "o\0", 0x1) = 1 0 19 6215/1: 349852 45 read(0x0, " \0", 0x1) = 1 0 20 6215/1: 54 39 write(0x2, " \0", 0x1) = 1 0 21 22In the above, the slow elapsed times for reads are due to the process context 23switching off the CPU while we wait for the next keystroke. For example, 24the second line shows an on-CPU time of 45 us and an elapsed time of 357210 us. 25In fact, the elapsed times are equal to the inter-keystroke delays. 26 27 28What about the writes? Their elapsed times are longer than the on-CPU times 29also. Did we context switch off for them too? ... Lets run a different demo, 30 31 # dtruss -eo date 32 ELAPSD CPU SYSCALL(args) = return 33 Mon Jul 25 21:41:40 EST 2005 34 44 23 resolvepath("/usr/bin/date\0", 0x80476CC, 0x3FF) = 13 0 35 10 1 sysconfig(0x6, 0xB25A1, 0xFEC1D444) = 4096 0 36 36 28 resolvepath("/usr/lib/ld.so.1\0", 0x80476CC, 0x3FF) = 12 0 37 18 9 xstat(0x2, 0x8047FEB, 0x8047AF8) = 0 0 38 25 16 open("/var/ld/ld.config\0", 0x0, 0x0) = -1 Err#2 39 27 18 xstat(0x2, 0xD27FBF38, 0x80473B0) = 0 0 40 17 9 resolvepath("/lib/libc.so.1\0", 0x8047438, 0x3FF) = 14 0 41 21 13 open("/lib/libc.so.1\0", 0x0, 0x0) = 3 0 42 30 22 mmap(0x10000, 0x1000, 0x5) = -763559936 0 43 15 6 mmap(0x10000, 0xCE000, 0x0) = -764411904 0 44 24 16 mmap(0xD2700000, 0xB5A45, 0x5) = -764411904 0 45 21 12 mmap(0xD27C6000, 0x5EB3, 0x3) = -763600896 0 46 18 9 mmap(0xD27CC000, 0x15C0, 0x3) = -763576320 0 47 14 5 munmap(0xD27B6000, 0x10000) = 0 0 48 186 176 memcntl(0xD2700000, 0x1B8D8, 0x4) = 0 0 49 17 7 close(0x3) = 0 0 50 [...] 51 52For every syscall, the elapsed time is around 10 us (microseconds) slower 53than the on-cpu time. These aren't micro context switches, this is due to 54DTrace slowing down the program! The more closely we measure something the 55more we effect it. (See Heisenberg's uncertainty principle). 56 57Ok, so for the above output we can tell that each elapsed time is around 10 us 58longer than it should be. That's fine, since it's fairly consistant and not 59a huge difference. This is an x86 server with a 867 MHz CPU. 60 61 62Now lets try the same on an Ultra 5 with a 360 MHz CPU, 63 64 # dtruss -eo date 65 ELAPSD CPU SYSCALL(args) = return 66 216 142 resolvepath("/usr/bin/date\0", 0xFFBFF338, 0x3FF) = 13 0 67 234 187 resolvepath("/usr/lib/ld.so.1\0", 0xFFBFF338, 0x3FF) = 12 0 68 113 67 stat("/usr/bin/date\0", 0xFFBFF818, 0xFFBFFFEB) = 0 0 69 136 90 open("/var/ld/ld.config\0", 0x0, 0x0) = -1 Err#2 70 107 61 stat("/opt/onbld/lib/libc.so.1\0", 0xFFBFF330, 0xFFBFF55C) = -1 Err#2 71 98 54 stat("/opt/SUNWspro/lib/libc.so.1\0", 0xFFBFF330, 0xFFBFF55C) = -1 Err#2 72 96 53 stat("/opt/SUNWmlib/lib/libc.so.1\0", 0xFFBFF330, 0xFFBFF55C) = -1 Err#2 73 97 54 stat("/usr/sfw/lib/libc.so.1\0", 0xFFBFF330, 0xFFBFF55C) = -1 Err#2 74 96 53 stat("/lib/libc.so.1\0", 0xFFBFF330, 0xFFBFF55C) = 0 0 75 134 92 resolvepath("/lib/libc.so.1\0", 0xFFBFEF30, 0x3FF) = 14 0 76 109 69 open("/lib/libc.so.1\0", 0x0, 0x0) = 3 0 77 177 132 mmap(0x10000, 0x2000, 0x5) = -12976128 0 78 [...] 79 80Now the time difference is around 40 us, and fairly consistant. 81 82 83This difference is find so long as we bear it in mind. Or, run DTrace 84on faster servers where the difference is much less. 85 86 87 88* The output appears shuffled? 89 90Read the answer to this in ALLsnoop_notes.txt. 91 92