RE: Netapp trunking - no performance gain

From: david byte (davidb@uptime.net)
Date: Wed Mar 30 2005 - 17:20:31 EST

  • Next message: Waters, G Scott RDECOM(THE LOG.SEC TEAM): "SMB signing and my F760 ..."

    The catch with trunking is that a single host will not see performance
    gains, unless they are load balancing across 2 nics and the filer is
    load balancing in the exact same manner. (not likely, but it can occur
    as aaron notes) Generally, trunking is used for load aggregation to
    provide higher performance to a larger number of clients. If you are
    looking for higher throughput that about 100MB/s you should be talking
    with your Network Appliance SE about assisting in the configuration and
    design of the system.

    David Byte
    Senior Systems Engineer
    918-461-9255
    davidb@uptime.net

    -----Original Message-----
    From: aaron hirsch
    Sent: Wednesday, March 30, 2005 3:53 PM
    To: toasters@mathworks.com
    Subject: RE: Netapp trunking - no performance gain

    > I'm trying to setup trunking on a netapp connected to a Cisco 3550
    >gigabit switch (OS vers.12 on the switch).

    Sorry no experience with this switch...

    > The odd thing is that when I setup two gigabit lines trunked together
    >from the netapp to the switch I saw no improvement in thoughtput
    >compared to a single gigabit line.

    Again, not familiar with this particular switch...however, I have setup
    an enviornment similar to what you described and did see performance
    increase instantly.

    Extreme Summit 200-48 with a FAS250. I created a single vif interface
    out of the 2 gige interfaces on the FAS250 and enabled sharing on the
    switch ports that they plug into. I then took two 10/100 cards on a
    server and bound them togther and enabled sharing on the ports for the
    switch too. I went from pushing about 10MB/sec from the server to
    20MB/sec. As this is well below the theoretical maximum for two gige
    ports trunked together, 250MB/sec (125MB/sec each), the problem is that
    pushing so little data makes it hard to see how effectively the gige
    trunk works.

    So, to answer your original question, yes I have done it; but utilizing
    10/100 cards doesn't do justice when testing performance on a gige
    device.



    This archive was generated by hypermail 2b29 : Wed Mar 30 2005 - 17:13:57 EST