Multiple NICS - Real Servers

John Cronin jsc3 at havoc.gtf.org
Thu Aug 9 04:26:44 BST 2001


> On Wed, Aug 08, 2001 at 01:21:27PM -0400, John Cronin wrote:
> > > 
> > > Joseph Mack wrote:
> > > > 1. channel bonding, where multiple NICs on a machine have the
> > > > same IP and MAC address and are used for higher throughput.
> > > > I read through 2yrs of messages on the beowulf mailing list
> > > > about this last night and couldn't see anywhere where people
> > > > allow this sort of setup to run after one card has failed.
> > > > I presume it's not doable.
> > > 
> > > It is not doable. If one of the NICs fails, every other packet will be 
> > > dropped.
> > 
> > Perhaps on one particular implementation this is true, but I don't
> > think you can say that in general, for all multi-NIC channel aggregation
> > setups, this is true.
> 
> Applogies for being vauge but I once saw a demonstration from Intel showing
> two 100Mb nics plugged into a switch. If both the Nics were available then
> both would be used, if one failed only one would be used. My understanding
> is that this system requires specific NICs and support on the switch but it
> did seem to work quite nicley. This was about 2 years ago and I haven't
> seen anything about it since.

It would be nice if it could work across multiple switches, so if a
single switch failed, you would not lose connectivity (I think the
adaptive failover can do this, but that does not improve bandwidth).

Even so, in a Cisco Catalyst chassis with multiple blades and redundant
supervisor cards (and power supplies, obviously), this might be a decent
HA/performance solution.  I'll have to look at the back of those big
Ciscos and see if they have multiple power cables.  I find it amusing
that a Sun 4500 can have four power supplies (and a 450 can have two),
but since they have a single power cord, one clumsy person can take
the system down quite easily.  In my experience, that kind of thing
happens far more often than equipment failure.  HA is all about
attention to details.

So, are there any Cisco Fast EtherChannel experts out there?  Can
FEC run across multiple switches, or at least across multiple Catalyst
blades?  I guess I can go look it up, but if somebody already knows,
I don't mind saving myself the trouble.

-- 
John Cronin
mailto: `echo NjsOc3 at SgtPfA.orMg | sed 's/[NOSPAM]//g'`




More information about the lvs-users mailing list