What could be the cause of this? - SUN

This is a discussion on What could be the cause of this? - SUN ; 4 servers on the same subnet. 1 Solaris 7 box, 1 Solaris 8 box and 2 Solaris 9 Sun boxes (T1, X1, V240 and V210 respectively). All servers are forced to 100 full duplex, so are the switch ports. No ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: What could be the cause of this?

  1. What could be the cause of this?

    4 servers on the same subnet. 1 Solaris 7 box, 1 Solaris 8 box and 2
    Solaris 9 Sun boxes (T1, X1, V240 and V210 respectively). All servers
    are forced to 100 full duplex, so are the switch ports. No errors or
    collisions on the switch ports, traceroute shows the exact same network
    path taken for all four servers, the round-trip times are identical.

    However, on FTP to a server on another site, downloading the same file,
    I get these results and I cannot figure out why:

    Solaris 7:
    5962683 bytes received in 38 seconds (152.58 Kbytes/s)

    Solaris 8:
    5962683 bytes received in 38 seconds (154.39 Kbytes/s)

    Solaris 9: (V240)
    5962683 bytes received in 20 seconds (297.85 Kbytes/s)

    Solaris 9: (V210)
    5962683 bytes received in 20 seconds (294.88 Kbytes/s)

    The Solaris 9 boxes yields almost twice the speed than the Solaris 7 and
    8 boxes!
    Now is there anything wrong on my nework or is the TCP/IP stack in
    Solaris 9 heavily tuned in comparison to Solaris 7 and 8? Or are the
    V210/240 servers simply shipped with better NIC's?

    All input and feedback is welcome, this problem is driving me crazy and
    I cannot figure out what's wrong!



    --
    Stig Bull
    | remove .no.spam from my email address to reply by mail |
    No animals were hurt or killed in the process of creating this
    electronic message. To reduce download time, this message is made of
    100% recycled bytes.

  2. Re: What could be the cause of this?

    On Mon, 19 Dec 2005, Stig Bull wrote:

    > Now is there anything wrong on my nework or is the TCP/IP stack in
    > Solaris 9 heavily tuned in comparison to Solaris 7 and 8? Or are the
    > V210/240 servers simply shipped with better NIC's?


    A bit of both. The V2[14]0 have onboard GBE, but the TCP/IP stack
    in SOlaris 9 is noticably better than on previous versions. When I
    upgraded my Ultra 1 web server to Solaris 9 (from 7 or 8, can't remember
    now), I literally had customers asking me what I'd done to improve
    the speed. I changed nothing except the OS: same HW, same version of
    apps, etc.

    --
    Rich Teer, SCNA, SCSA, OpenSolaris CAB member

    . * * . * .* .
    . * . .*
    President, * . . /\ ( . . *
    Rite Online Inc. . . / .\ . * .
    .*. / * \ . .
    . /* o \ .
    Voice: +1 (250) 979-1638 * '''||''' .
    URL: http://www.rite-group.com/rich ******************

  3. Re: What could be the cause of this?

    Stig Bull writes:
    >
    >However, on FTP to a server on another site, downloading the same file,
    >I get these results and I cannot figure out why:
    >


    In Solaris 8 and previous versions, the ftp client or server (can't
    remember which) limits its window to 24k packets. I don't know if
    that's true of the ftp in Solaris 9.

    Install TTCP and use it to test tcp throughput, not ftp.

    -Greg
    --
    Do NOT reply via e-mail.
    Reply in the newsgroup.

  4. Re: What could be the cause of this?

    In article ,
    rich.teer@rite-group.com says...
    > On Mon, 19 Dec 2005, Stig Bull wrote:
    >
    > > Now is there anything wrong on my nework or is the TCP/IP stack in
    > > Solaris 9 heavily tuned in comparison to Solaris 7 and 8? Or are the
    > > V210/240 servers simply shipped with better NIC's?

    >
    > A bit of both. The V2[14]0 have onboard GBE, but the TCP/IP stack
    > in SOlaris 9 is noticably better than on previous versions. When I
    > upgraded my Ultra 1 web server to Solaris 9 (from 7 or 8, can't remember
    > now), I literally had customers asking me what I'd done to improve
    > the speed. I changed nothing except the OS: same HW, same version of
    > apps, etc.


    I found the solution, by setting the following parameters in Solaris 7
    identical to Solaris 9 (ndd /dev/tcp), networking speed jumped from
    60kb/s to over 300kb/s:

    tcp_xmit_hiwat
    tcp_xmit_lowat
    tcp_recv_hiwat

    --
    Stig Bull
    | remove .no.spam from my email address to reply by mail |
    No animals were hurt or killed in the process of creating this
    electronic message. To reduce download time, this message is made of
    100% recycled bytes.

+ Reply to Thread