What does this sar result mean? - Unix

This is a discussion on What does this sar result mean? - Unix ; Hi, I have get a recent report from my machine and there is something I do not understand. Please take a look at the result first: tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util 143.93 1534.74 1342.01 19.99 18.49 128.43 ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: What does this sar result mean?

  1. What does this sar result mean?

    Hi,
    I have get a recent report from my machine and there is something I
    do not understand.
    Please take a look at the result first:

    tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util
    143.93 1534.74 1342.01 19.99 18.49 128.43 3.29 47.41
    120.24 1424.39 1054.50 20.62 14.47 120.33 3.41 41.00
    102.01 1270.69 976.18 22.03 12.87 126.20 3.80 38.79
    81.82 947.90 765.50 20.94 8.94 109.30 4.07 33.27
    89.55 660.23 730.50 15.53 7.76 86.61 3.80 34.01
    182.51 2088.51 1199.39 18.01 16.24 89.00 3.16 57.61
    73.11 1207.13 583.24 24.49 6.22 84.67 4.13 30.16
    56.75 866.81 504.46 24.17 4.06 72.02 4.71 26.73
    48.33 397.68 439.40 17.32 3.35 69.31 5.05 24.38
    46.44 490.52 430.11 19.82 3.43 73.84 4.94 22.94
    41.16 330.09 367.70 16.95 2.86 69.38 5.36 22.05
    408.19 13068.02 1939.98 36.77 7.25 17.75 2.21 90.23
    469.36 5933.15 1763.28 16.40 8.07 17.20 2.13 99.89
    463.78 5695.35 1419.88 15.34 5.76 12.42 2.15 99.86
    446.36 5535.90 1712.86 16.24 6.81 15.25 2.24 99.83
    449.05 5166.26 1424.83 14.68 6.43 14.31 2.22 99.89
    460.23 5596.55 1585.21 15.60 6.44 14.00 2.17 99.85
    454.00 5361.55 1518.18 15.15 7.34 16.18 2.20 99.88
    576.51 7397.60 1870.47 16.08 9.02 15.64 1.73 99.91
    562.26 8555.19 2776.98 20.15 9.28 16.50 1.78 99.92
    561.50 9001.87 3462.99 22.20 11.94 21.27 1.78 99.95
    502.35 9393.50 4081.07 26.82 12.17 24.22 1.99 99.98

    I am sorry that the format is still a little confused although I have
    make some change to it to be more suitable for news group. Any way, I
    think it is now readable.

    Above is I/O report from 00:00 to 06:00 or some time in my machine.
    There are about two main status in the report, in the first half the
    whole load is very normal and the %util is about 40%. Suddenly, the
    %util is up to 99% and the whole disk is in high load. I know, in that
    time I run some I/O sensive task. But I can not understand that when the
    %util is low the await/svctm/avgqu-sz is very high, but when the %util
    is high the await/svctm/avgqu-sz is very low. I do not understand why?

    I guess that maybe this has something to do with the way Linux deal with
    I/O. But I can not figure out the real reason, could you please give me
    some detailed explaination? Thanks in advance!

    Regards!
    Bo

  2. Re: What does this sar result mean?

    Bo Yang wrote:
    >
    > I have get a recent report from my machine and there is something I
    > do not understand.
    > Please take a look at the result first:
    >
    > tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util
    > 143.93 1534.74 1342.01 19.99 18.49 128.43 3.29 47.41
    > 120.24 1424.39 1054.50 20.62 14.47 120.33 3.41 41.00
    > 102.01 1270.69 976.18 22.03 12.87 126.20 3.80 38.79
    > 81.82 947.90 765.50 20.94 8.94 109.30 4.07 33.27
    > 89.55 660.23 730.50 15.53 7.76 86.61 3.80 34.01
    > 182.51 2088.51 1199.39 18.01 16.24 89.00 3.16 57.61
    > 73.11 1207.13 583.24 24.49 6.22 84.67 4.13 30.16
    > 56.75 866.81 504.46 24.17 4.06 72.02 4.71 26.73
    > 48.33 397.68 439.40 17.32 3.35 69.31 5.05 24.38
    > 46.44 490.52 430.11 19.82 3.43 73.84 4.94 22.94
    > 41.16 330.09 367.70 16.95 2.86 69.38 5.36 22.05
    > 408.19 13068.02 1939.98 36.77 7.25 17.75 2.21 90.23
    > 469.36 5933.15 1763.28 16.40 8.07 17.20 2.13 99.89
    > 463.78 5695.35 1419.88 15.34 5.76 12.42 2.15 99.86
    > 446.36 5535.90 1712.86 16.24 6.81 15.25 2.24 99.83
    > 449.05 5166.26 1424.83 14.68 6.43 14.31 2.22 99.89
    > 460.23 5596.55 1585.21 15.60 6.44 14.00 2.17 99.85
    > 454.00 5361.55 1518.18 15.15 7.34 16.18 2.20 99.88
    > 576.51 7397.60 1870.47 16.08 9.02 15.64 1.73 99.91
    > 562.26 8555.19 2776.98 20.15 9.28 16.50 1.78 99.92
    > 561.50 9001.87 3462.99 22.20 11.94 21.27 1.78 99.95
    > 502.35 9393.50 4081.07 26.82 12.17 24.22 1.99 99.98
    >
    > I am sorry that the format is still a little confused although I have
    > make some change to it to be more suitable for news group. Any way, I
    > think it is now readable.
    >
    > Above is I/O report from 00:00 to 06:00 or some time in my machine.
    > There are about two main status in the report, in the first half the
    > whole load is very normal and the %util is about 40%. Suddenly, the
    > %util is up to 99% and the whole disk is in high load. I know, in that
    > time I run some I/O sensive task. But I can not understand that when the
    > %util is low the await/svctm/avgqu-sz is very high, but when the %util
    > is high the await/svctm/avgqu-sz is very low. I do not understand why?


    Before the backups started the requests were random so the
    read-ahead function didn't help. While backups ran the
    read-ahead function worked very well for the backups and that
    dominated the stats.

    > I guess that maybe this has something to do with the way Linux deal with
    > I/O. But I can not figure out the real reason, could you please give me
    > some detailed explaination? Thanks in advance!


    Both arrays and disks do read-ahead these days in addition to
    device drivers.


  3. Re: What does this sar result mean?

    Doug Freyburger :
    > Bo Yang wrote:
    >> I have get a recent report from my machine and there is something I
    >> do not understand.
    >> Please take a look at the result first:
    >>
    >> tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util
    >> 143.93 1534.74 1342.01 19.99 18.49 128.43 3.29 47.41
    >> 120.24 1424.39 1054.50 20.62 14.47 120.33 3.41 41.00
    >> 102.01 1270.69 976.18 22.03 12.87 126.20 3.80 38.79
    >> 81.82 947.90 765.50 20.94 8.94 109.30 4.07 33.27
    >> 89.55 660.23 730.50 15.53 7.76 86.61 3.80 34.01
    >> 182.51 2088.51 1199.39 18.01 16.24 89.00 3.16 57.61
    >> 73.11 1207.13 583.24 24.49 6.22 84.67 4.13 30.16
    >> 56.75 866.81 504.46 24.17 4.06 72.02 4.71 26.73
    >> 48.33 397.68 439.40 17.32 3.35 69.31 5.05 24.38
    >> 46.44 490.52 430.11 19.82 3.43 73.84 4.94 22.94
    >> 41.16 330.09 367.70 16.95 2.86 69.38 5.36 22.05
    >> 408.19 13068.02 1939.98 36.77 7.25 17.75 2.21 90.23
    >> 469.36 5933.15 1763.28 16.40 8.07 17.20 2.13 99.89
    >> 463.78 5695.35 1419.88 15.34 5.76 12.42 2.15 99.86
    >> 446.36 5535.90 1712.86 16.24 6.81 15.25 2.24 99.83
    >> 449.05 5166.26 1424.83 14.68 6.43 14.31 2.22 99.89
    >> 460.23 5596.55 1585.21 15.60 6.44 14.00 2.17 99.85
    >> 454.00 5361.55 1518.18 15.15 7.34 16.18 2.20 99.88
    >> 576.51 7397.60 1870.47 16.08 9.02 15.64 1.73 99.91
    >> 562.26 8555.19 2776.98 20.15 9.28 16.50 1.78 99.92
    >> 561.50 9001.87 3462.99 22.20 11.94 21.27 1.78 99.95
    >> 502.35 9393.50 4081.07 26.82 12.17 24.22 1.99 99.98
    >>
    >> I am sorry that the format is still a little confused although I have
    >> make some change to it to be more suitable for news group. Any way, I
    >> think it is now readable.
    >>
    >> Above is I/O report from 00:00 to 06:00 or some time in my machine.
    >> There are about two main status in the report, in the first half the
    >> whole load is very normal and the %util is about 40%. Suddenly, the
    >> %util is up to 99% and the whole disk is in high load. I know, in that
    >> time I run some I/O sensive task. But I can not understand that when the
    >> %util is low the await/svctm/avgqu-sz is very high, but when the %util
    >> is high the await/svctm/avgqu-sz is very low. I do not understand why?

    >
    > Before the backups started the requests were random so the
    > read-ahead function didn't help. While backups ran the
    > read-ahead function worked very well for the backups and that
    > dominated the stats.
    >
    >> I guess that maybe this has something to do with the way Linux deal with
    >> I/O. But I can not figure out the real reason, could you please give me
    >> some detailed explaination? Thanks in advance!

    >
    > Both arrays and disks do read-ahead these days in addition to
    > device drivers.
    >

    You mean the disk performance will be improved when I/O are sequent? Thanks!

    Regards!
    Bo

+ Reply to Thread