Can eri network interface run at 1000 fdx? - SUN

This is a discussion on Can eri network interface run at 1000 fdx? - SUN ; Hi All, Does anyone know that can eri network interface run at 1000 fdx? If it does or doesn't, please provide me a link/url which tells me about hw architecture of eri interface so that I can hard-code a port ...

+ Reply to Thread
Results 1 to 9 of 9

Thread: Can eri network interface run at 1000 fdx?

  1. Can eri network interface run at 1000 fdx?

    Hi All,

    Does anyone know that can eri network interface run at 1000 fdx? If
    it does or doesn't, please provide me a link/url which tells me about
    hw architecture of eri interface so that I can hard-code a port on a
    switch correctly? I think it can only run at 100 fdx, but not sure.
    Any helps are appreciated.

    TIA,
    -Chris


  2. Re: Can eri network interface run at 1000 fdx?

    "lazyboy_2k@yahoo.com" writes:

    >Does anyone know that can eri network interface run at 1000 fdx? If
    >it does or doesn't, please provide me a link/url which tells me about
    >hw architecture of eri interface so that I can hard-code a port on a
    >switch correctly? I think it can only run at 100 fdx, but not sure.
    >Any helps are appreciated.


    Set your switch to autoconf.

    It does 10/100 hdx/fdx.

    If you set the switch to 100-full, the system will negotiate
    100-*half* [as required by the standard ] and you will be very unhappy.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

  3. Re: Can eri network interface run at 1000 fdx?

    lazyboy_2k@yahoo.com wrote:
    > Hi All,
    >
    > Does anyone know that can eri network interface run at 1000 fdx? If
    > it does or doesn't, please provide me a link/url which tells me about
    > hw architecture of eri interface so that I can hard-code a port on a
    > switch correctly? I think it can only run at 100 fdx, but not sure.
    > Any helps are appreciated.
    >
    > TIA,
    > -Chris
    >


    Setting speed and duplex on a switch port is usually the Wrong Thing To
    Do! Autonegotiation works best if it works. There have been cases with
    older NICs and older switches where autonegotiation failed due to
    ambiguities in the standard. Reasonably modern hardware should
    autonegotiate without problems.

    FWIW the eri is 10/100 fdx. If you want gigabit, you will have to
    purchase and install hardware capable of gigabit speeds.


  4. Re: Can eri network interface run at 1000 fdx?

    On May 15, 9:27 pm, Casper H.S. Dik wrote:

    > If you set the switch to 100-full, the system will negotiate
    > 100-*half* [as required by the standard ] and you will be very unhappy.


    Can things be configured so that this will work (in the sense of:
    arrive at 100/full or fail)? I realise you can obviously force the
    machine as well as the switch, but is it also legal to set things so
    that one or both of the partners will negotiate, but will only
    negotiate fdx. It looks to me like empirically you might be able to
    from a Solaris box - presumably I could try setting adv_100hdx_cap &
    similar things to false. But I'm wondering if it's legal to do that
    for ethernet: I can see it might not be since half-duplex is some kind
    of fallback thing.

    This isn't just idle curiosity. I frequently come across people who
    configure switches to 100/full (or 1000/full) with no negotiation on
    the grounds that they want to ensure 100/full, and obviously don't
    understand that what they'll actually achieve is either something much
    worse than 100/half or endless pain for people running the machines
    connected to the switch, and usually both. If I could go back to them
    and say "set it to negotiate but only accept 100/full" they might
    accept that. And not only that it would achieve what they want which
    is for the link to not come up at all if it both partners can't agree
    on fdx: that's a lot better than arriving at a mismatch, because it is
    much easier to detect.

    (Of course what will actually happen is that they won't listen because
    I'm not a network person and therefore can't possibly understand
    anything.)

    --tim


  5. Re: Can eri network interface run at 1000 fdx?

    Tim Bradshaw writes:

    >On May 15, 9:27 pm, Casper H.S. Dik wrote:


    >> If you set the switch to 100-full, the system will negotiate
    >> 100-*half* [as required by the standard ] and you will be very unhappy.


    >Can things be configured so that this will work (in the sense of:
    >arrive at 100/full or fail)? I realise you can obviously force the
    >machine as well as the switch, but is it also legal to set things so
    >that one or both of the partners will negotiate, but will only
    >negotiate fdx. It looks to me like empirically you might be able to
    >from a Solaris box - presumably I could try setting adv_100hdx_cap &
    >similar things to false. But I'm wondering if it's legal to do that
    >for ethernet: I can see it might not be since half-duplex is some kind
    >of fallback thing.


    I'm not sure how this works to that particular level of detail;
    or whether the switch even has a mode that is "autonegotiate 100-full only."

    I believe Solaris can, but I can't say without testing.
    Mileage will vary by switch type, vendor and firmware revision.

    >This isn't just idle curiosity. I frequently come across people who
    >configure switches to 100/full (or 1000/full) with no negotiation on
    >the grounds that they want to ensure 100/full, and obviously don't
    >understand that what they'll actually achieve is either something much
    >worse than 100/half or endless pain for people running the machines
    >connected to the switch, and usually both. If I could go back to them
    >and say "set it to negotiate but only accept 100/full" they might
    >accept that. And not only that it would achieve what they want which
    >is for the link to not come up at all if it both partners can't agree
    >on fdx: that's a lot better than arriving at a mismatch, because it is
    >much easier to detect.


    >(Of course what will actually happen is that they won't listen because
    >I'm not a network person and therefore can't possibly understand
    >anything.)


    Quite :-)

    I think that in our server rooms we used to have most things set to
    100/full because auto negotiation didn't work at some time in the past.

    So I debugged a lot of "poor network performance issues" were the
    solution was to simply "fix the switch".

    FOrtunately, we now have mostly SunRAYs so the switches just have to
    be set to autonegotiation.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

  6. Re: Can eri network interface run at 1000 fdx?

    Tim Bradshaw wrote:
    > On May 15, 9:27 pm, Casper H.S. Dik wrote:
    >
    >
    >>If you set the switch to 100-full, the system will negotiate
    >>100-*half* [as required by the standard ] and you will be very unhappy.

    >
    >
    > Can things be configured so that this will work (in the sense of:
    > arrive at 100/full or fail)? I realise you can obviously force the
    > machine as well as the switch, but is it also legal to set things so
    > that one or both of the partners will negotiate, but will only
    > negotiate fdx. It looks to me like empirically you might be able to
    > from a Solaris box - presumably I could try setting adv_100hdx_cap &
    > similar things to false. But I'm wondering if it's legal to do that
    > for ethernet: I can see it might not be since half-duplex is some kind
    > of fallback thing.
    >
    > This isn't just idle curiosity. I frequently come across people who
    > configure switches to 100/full (or 1000/full) with no negotiation on
    > the grounds that they want to ensure 100/full, and obviously don't
    > understand that what they'll actually achieve is either something much
    > worse than 100/half or endless pain for people running the machines
    > connected to the switch, and usually both. If I could go back to them
    > and say "set it to negotiate but only accept 100/full" they might
    > accept that. And not only that it would achieve what they want which
    > is for the link to not come up at all if it both partners can't agree
    > on fdx: that's a lot better than arriving at a mismatch, because it is
    > much easier to detect.
    >
    > (Of course what will actually happen is that they won't listen because
    > I'm not a network person and therefore can't possibly understand
    > anything.)
    >
    > --tim
    >


    You can set the Solaris end using ndd. You set the "cap" (capacity?) to
    say, in effect, "I only know how to do fdx". Then it negotiates, as
    required by the standard, and the negotiation should result in fdx as
    long as both ends are capable.

    Actually, this screwing around should NOT be necessary. If both ends
    are capable of 100/fdx and both are set to autonegotiate, they SHOULD
    negotiate and agree on 100/fdx. This is not necessarily true for some
    very old (1990's) hardware due to ambiguities in the standard.

    And even the network people have to listen to the people who pay the
    bills! Bear in mind that, if you go this route they will probably hate
    you forever! Especially if it turns out that you were right! This is
    not necessarily a bad thing; I was there when they "came out of the
    closet!" ;-)


  7. Re: Can eri network interface run at 1000 fdx?

    Casper H.S. Dik wrote:
    > If you set the switch to 100-full, the system will negotiate
    > 100-*half* [as required by the standard ] and you will be very
    > unhappy.


    Well, it won't really "negotiate" right? Autoneg will "fail" because
    the side hardcoded to 100FD will not autoneg, and when autoneg "fails"
    the spec says the side attempting autoneg must go to half-duplex.

    Some boilerplate I trot-out from time to time:

    How 100Base-T Autoneg is supposed to work:

    When both sides of the link are set to autoneg, they will "negotiate"
    the duplex setting and select full-duplex if both sides can do
    full-duplex.

    If one side is hardcoded and not using autoneg, the autoneg process
    will "fail" and the side trying to autoneg is required by spec to use
    half-duplex mode.

    If one side is using half-duplex, and the other is using full-duplex,
    sorrow and woe is the usual result.

    So, the following table shows what will happen given various settings
    on each side:

    Auto Half Full

    Auto Happiness Lucky Sorrow

    Half Lucky Happiness Sorrow

    Full Sorrow Sorrow Happiness

    Happiness means that there is a good shot of everything going well.
    Lucky means that things will likely go well, but not because you did
    anything correctly Sorrow means that there _will_ be a duplex
    mis-match.

    When there is a duplex mismatch, on the side running half-duplex you
    will see various errors and probably a number of _LATE_ collisions
    ("normal" collisions don't count here). On the side running
    full-duplex you will see things like FCS errors. Note that those
    errors are not necessarily conclusive, they are simply indicators.

    Further, it is important to keep in mind that a "clean" ping (or the
    like - eg "linkloop" or default netperf TCP_RR) test result is
    inconclusive here - a duplex mismatch causes lost traffic _only_ when
    both sides of the link try to speak at the same time. A typical ping
    test, being synchronous, one at a time request/response, never tries
    to have both sides talking at the same time.

    Finally, when/if you migrate to 1000Base-T, everything has to be set
    to auto-neg anyway.


    --
    oxymoron n, Hummer H2 with California Save Our Coasts and Oceans plates
    these opinions are mine, all mine; HP might not want them anyway...
    feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

  8. Re: Can eri network interface run at 1000 fdx?

    On 2007-05-16 16:03:54 +0100, "Richard B. Gilbert"
    said:

    > Actually, this screwing around should NOT be necessary. If both ends
    > are capable of 100/fdx and both are set to autonegotiate, they SHOULD
    > negotiate and agree on 100/fdx. This is not necessarily true for some
    > very old (1990's) hardware due to ambiguities in the standard.


    I realise that, but I'm wondering if some approach short of "learn how
    to do your job" might work.

    >
    > And even the network people have to listen to the people who pay the
    > bills! Bear in mind that, if you go this route they will probably hate
    > you forever!


    They already do...


  9. Re: Can eri network interface run at 1000 fdx?

    Rick Jones writes:

    >Casper H.S. Dik wrote:
    >> If you set the switch to 100-full, the system will negotiate
    >> 100-*half* [as required by the standard ] and you will be very
    >> unhappy.


    >Well, it won't really "negotiate" right? Autoneg will "fail" because
    >the side hardcoded to 100FD will not autoneg, and when autoneg "fails"
    >the spec says the side attempting autoneg must go to half-duplex.


    Well, yes.

    The SPARC system I used, though, has a few knobs that can't really make
    it do "autonegotiate only to 100Mbps full".

    What you can set is:

    - supported modes 10h, 10f, 100h, 100f
    - autoneg on/off

    with autoneg set to on or off but the only supported mode 100/full,
    you'll always get 100 full.

    But you can observe that in the one case the switch will fallback to
    100/half as it should and in the "autoneg on" case the switch will
    "negotiate" 100/full.

    Technically if it fails it should fall back to 100/half
    but you can tell it not too.

    Casper

+ Reply to Thread