Redundant Linked Edge Devices - Some theory on Dual Homing, please

Discussion in 'Cisco' started by Chris Harvey, Jan 12, 2004.

  1. Chris Harvey

    Chris Harvey Guest

    Hello

    I am looking for some help regarding best practices for redundant linked
    edge devices (PCs etc).

    Basically, a customer would like dual homed PCs (Clients, not servers!)
    linked to separate Cisco switches in a classical Access/Distribution/Core
    architecture. This is to provide extra special high availability to certain
    users.

    My instinct cries out that this is a bad idea. I can visualise lots of
    problems if the devices connect at layer 2, and even more if they connect at
    layer 3 (yuck).

    What I am looking at is either reading reference on this subject, possible
    problems I could expect or some anecdotal evidence from people who have
    tried this in the past (citing vendors etc).

    I need to put together a section in a customer proposal to disuade them of
    this path and I'd appreciate some help.

    Cheers

    Chris
     
    Chris Harvey, Jan 12, 2004
    #1
    1. Advertising

  2. Chris Harvey

    hdu Guest

    My company use Intel Server Pro adapter and use their solution (iANS) which
    support NT, Win2k, Linux, ...

    You can goto Intel's web page for more information.

    "Chris Harvey" <> ¦b¶l¥ó news:btu87m$qg1$ ¤¤
    ¼¶¼g...
    > Hello
    >
    > I am looking for some help regarding best practices for redundant linked
    > edge devices (PCs etc).
    >
    > Basically, a customer would like dual homed PCs (Clients, not servers!)
    > linked to separate Cisco switches in a classical Access/Distribution/Core
    > architecture. This is to provide extra special high availability to

    certain
    > users.
    >
    > My instinct cries out that this is a bad idea. I can visualise lots of
    > problems if the devices connect at layer 2, and even more if they connect

    at
    > layer 3 (yuck).
    >
    > What I am looking at is either reading reference on this subject, possible
    > problems I could expect or some anecdotal evidence from people who have
    > tried this in the past (citing vendors etc).
    >
    > I need to put together a section in a customer proposal to disuade them of
    > this path and I'd appreciate some help.
    >
    > Cheers
    >
    > Chris
    >
    >
    >
     
    hdu, Jan 12, 2004
    #2
    1. Advertising

  3. Chris Harvey

    Chris Harvey Guest

    "hdu" <> wrote in message news:4002cda1$...
    > My company use Intel Server Pro adapter and use their solution (iANS)

    which
    > support NT, Win2k, Linux, ...
    >
    > You can goto Intel's web page for more information.


    But surely not at the user level. Typically (for a large campus) a designer
    would use layer 2 at the user access layers, with layer 3 uplinks. The
    thought of hundreds of devices recalculating spanning tree makes me shudder.
    Add to that a proprietary algorithm for "application level healing", and I
    believe this to be a recipe for disaster.

    Any more comments?

    Chris




    >
    > "Chris Harvey" <> ¦b¶l¥ó news:btu87m$qg1$

    ¤¤
    > ¼¶¼g...
    > > Hello
    > >
    > > I am looking for some help regarding best practices for redundant linked
    > > edge devices (PCs etc).
    > >
    > > Basically, a customer would like dual homed PCs (Clients, not servers!)
    > > linked to separate Cisco switches in a classical

    Access/Distribution/Core
    > > architecture. This is to provide extra special high availability to

    > certain
    > > users.
    > >
    > > My instinct cries out that this is a bad idea. I can visualise lots of
    > > problems if the devices connect at layer 2, and even more if they

    connect
    > at
    > > layer 3 (yuck).
    > >
    > > What I am looking at is either reading reference on this subject,

    possible
    > > problems I could expect or some anecdotal evidence from people who have
    > > tried this in the past (citing vendors etc).
    > >
    > > I need to put together a section in a customer proposal to disuade them

    of
    > > this path and I'd appreciate some help.
    > >
    > > Cheers
    > >
    > > Chris
    > >
    > >
    > >

    >
    >
     
    Chris Harvey, Jan 12, 2004
    #3
  4. Hi,

    You do not necessarily have to include spanning-tree to get redundant
    layer-2 solutions. The Intel PROset software allows you to "team"-together
    two network interfaces using several techniques. One of them is the FEC/GEC
    which you can't use because you'd like to connect to two seperate switches,
    but there is another one that implements a kind of heart-beat between the
    two nics to find out if one has a problem or not. If there is a problem the
    client-software switches to the other nic, transferring mac-addresses (if I
    recall correctly). There is no spanning-tree involved in this, no special
    switch-configuration either.

    Erik

    "Chris Harvey" <> wrote in message
    news:btujkt$7gb$...
    >
    > "hdu" <> wrote in message news:4002cda1$...
    > > My company use Intel Server Pro adapter and use their solution (iANS)

    > which
    > > support NT, Win2k, Linux, ...
    > >
    > > You can goto Intel's web page for more information.

    >
    > But surely not at the user level. Typically (for a large campus) a

    designer
    > would use layer 2 at the user access layers, with layer 3 uplinks. The
    > thought of hundreds of devices recalculating spanning tree makes me

    shudder.
    > Add to that a proprietary algorithm for "application level healing", and I
    > believe this to be a recipe for disaster.
    >
    > Any more comments?
    >
    > Chris
    >
    >
    >
    >
    > >
    > > "Chris Harvey" <> ¦b¶l¥ó

    news:btu87m$qg1$
    > ¤¤
    > > ¼¶¼g...
    > > > Hello
    > > >
    > > > I am looking for some help regarding best practices for redundant

    linked
    > > > edge devices (PCs etc).
    > > >
    > > > Basically, a customer would like dual homed PCs (Clients, not

    servers!)
    > > > linked to separate Cisco switches in a classical

    > Access/Distribution/Core
    > > > architecture. This is to provide extra special high availability to

    > > certain
    > > > users.
    > > >
    > > > My instinct cries out that this is a bad idea. I can visualise lots

    of
    > > > problems if the devices connect at layer 2, and even more if they

    > connect
    > > at
    > > > layer 3 (yuck).
    > > >
    > > > What I am looking at is either reading reference on this subject,

    > possible
    > > > problems I could expect or some anecdotal evidence from people who

    have
    > > > tried this in the past (citing vendors etc).
    > > >
    > > > I need to put together a section in a customer proposal to disuade

    them
    > of
    > > > this path and I'd appreciate some help.
    > > >
    > > > Cheers
    > > >
    > > > Chris
    > > >
    > > >
    > > >

    > >
    > >

    >
    >
     
    Erik Tamminga, Jan 12, 2004
    #4
  5. >> "hdu" <> wrote in message news:4002cda1$...
    >> >
    >> > You can goto Intel's web page for more information.

    >>
    >> But surely not at the user level. Typically (for a large campus) a designer
    >> would use layer 2 at the user access layers, with layer 3 uplinks. The
    >> thought of hundreds of devices recalculating spanning tree makes me shudder.
    >> Add to that a proprietary algorithm for "application level healing", and I
    >> believe this to be a recipe for disaster.
    >>
    >> Any more comments?
    >>
    >> Chris
    >> >
    >> > "Chris Harvey" <> ¦b¶l¥ó
    >> > > Hello
    >> > >
    >> > > I am looking for some help regarding best practices for redundant linked
    >> > > edge devices (PCs etc).
    >> > >
    >> > > Basically, a customer would like dual homed PCs (Clients, not servers!)
    >> > > linked to separate Cisco switches in a classical Access/Distribution/Core
    >> > > architecture. This is to provide extra special high availability to certain
    >> > > users.
    >> > >
    >> > > My instinct cries out that this is a bad idea. I can visualise lots of
    >> > > problems if the devices connect at layer 2, and even more if they connect
    >> > > at layer 3 (yuck).
    >> > >
    >> > > What I am looking at is either reading reference on this subject, possible
    >> > > problems I could expect or some anecdotal evidence from people who have
    >> > > tried this in the past (citing vendors etc).
    >> > >
    >> > > I need to put together a section in a customer proposal to disuade
    >> > > them of this path and I'd appreciate some help.
    >> > >
    >> > > Cheers
    >> > >
    >> > > Chris


    I devoted a chapter to this problem in my book, and the bottom line
    is that there are a number of approaches, all of which work under
    some circumstances and fail under others. In practice, I have found
    it a less than satisfactory approach to improving the availability
    of clients, as a LAN link should NOT be a major source of down time
    in a well designed network. Compare the failure rate of a reasonable
    quality switch to that of a consumer grade PC.

    Bottom line, unless your client is already at the 4 to 5 nines level of
    availability, the time and money would be better spent upgrading the
    infrastructure. For example, do they have a UPS in the wiring closet
    and for each "important" client system, backup generator(s) to keep the
    HVAC working in the computer room, routine monitoring and testing of
    all redundancy already in place, a tested disaster recovery plan, etc.)

    On the other hand, if they already have a solid infrastructure in
    place, you should be able to upgrade each of the clients to a location
    independent thin client served by a redundant cluster :) to complement
    the multihoming provided by the upgraded network! Don't forget to
    provide physical diversity in your redundant solutions and set proper
    expectations for the management costs to keep it running!

    Good luck and have fun!
    --
    Vincent C Jones, Consultant Expert advice and a helping hand
    Networking Unlimited, Inc. for those who want to manage and
    Tenafly, NJ Phone: 201 568-7810 control their networking destiny
    http://www.networkingunlimited.com
     
    Vincent C Jones, Jan 12, 2004
    #5
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. David MacKay
    Replies:
    0
    Views:
    751
    David MacKay
    Nov 5, 2003
  2. Replies:
    12
    Views:
    4,637
    Hansang Bae
    Aug 31, 2006
  3. happy.ppp

    Prevent\Detect Dual Homing

    happy.ppp, Aug 27, 2006, in forum: Wireless Networking
    Replies:
    2
    Views:
    689
    happy.ppp
    Aug 28, 2006
  4. Ian

    Lenovo ThinkPad EDGE 13: Bleeding Edge

    Ian, Feb 28, 2011, in forum: Front Page News
    Replies:
    0
    Views:
    1,229
  5. Dale

    string theory, M-theory etc.

    Dale, Nov 7, 2013, in forum: Digital Photography
    Replies:
    1
    Views:
    167
    Martin Leese
    Nov 7, 2013
Loading...

Share This Page