RE: Measuring failover time from a link failure

["David Gillett" <gillettdavid () fhda edu>]

  There are at least two different technologies on your list,
neither of which necessarily involves anything normally called
"failover". 

  An Ethernet trunk uses some form of packet tagging to carry
traffic for multiple VLANs over a single physical connection.
There is no inherent failover -- if the physical link goes 
dead, all of the VLANs are disconnected over that link.
  It is common for trunks to be used in an environment where
OSPF or EIGRP or STP will be used to redirect traffic over
some redundant physical path, but any or all of those could
equally be used in an environment without trunking, so your
question should be about them and not about trunking itself.

  I believe all of the other terms you've used refer to treating
a group of parallel connections as a single high-bandwidth link.
Obviously any such scheme needs a way to handle outages on individual 
physical connections within the group, but the rest of the world
sees this as fluctuation of the available bandwidth on the
aggregated link, and not as failover to a different link.
  Note that if all the members of the aggregate flow through the
same conduit, they can all be severed by a single "backhoe interrupt",
requiring failover to an alternate path/route -- see OSPF etc above.

David Gillett



> -----Original Message-----
> From: Matthias Merk [mailto:macem99 () gmail com] 
> Sent: Thursday, December 20, 2007 3:06 AM
> To: security-basics () securityfocus com
> Subject: Measuring failover time from a link failure
>
> Hi
>
> I'm about to evaluate some Core Routers (Catalyst/Passport 
> class) so my question is, does anyone know of a tool/script 
> or a method to measure failover times of a "Ethernet trunk", 
> "NIC teaming", "port teaming", "port trunking", 
> "EtherChannel", "Multi-Link Trunking (MLT)"
> ... whatever you want to call it.
>
> Googling for "measuring link failover time" or something like 
> that didnt bring up much good.
>
> Thanks in advance!
> Matthias