lkml.org 
[lkml]   [1999]   [Apr]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: fast EtherChannel support?
Alan Cox wrote:
>
> > Anybody know if fast etherchannel support is available for Linux? This is
> > the Cisco protocol where two to four ethernet interfaces can be bound
> > together as one logical high-speed interface.
>
> Ask cisco for the protocol 8)

I did. Last time Cisco's switching product manager was
in Australia I asked for the Cisco ISL (ie: VLAN) and
etherchannel protocol specs.

The ISL encoding spec is public, the spec for their VTP
control protocol is not. I asked why not and basically
their legal people want to tie up patents first. I asked
would the free software community get a free license to
the patents, he said yes.

Of course, none of this is too relevant, as anyone in
their right mind would use IEEE rather than Cisco VLANs.
Then again, implementing both IEEE and ISL VLANs would
force a suitable degree of abstraction into any ethernet
VLAN code.

Also, in the tradition of "a 486 is a terrible thing to
waste" you can pick up ISL-only switches pretty cheaply
second-hand.


The EtherChannel encoding is:

> Our algorithm (currently) is
> (destination_ethernet_address AND 000.0000.0003) XOR
> (source_ethernet_address AND 0000.0000.0003)

You then round-robin the resulting eight values through
however many ports are in the channel.

The scheme has a few flaws, but remember that it was
designed for hardware implementation and modulo prime
schemes use lots of silicon.


No traffic policing is done on received packets. So in
practice, EtherChannel will happily interwork with other
hashing schemes.

Cisco does have a control protocol for EtherChannel, mainly
to allow autoconfiguration and self-healing on links that
show carrier but are actually broken (as usually happens
when you insert some 100BaseTX/100BaseFX repeaters).

This protocol is not available to the public. It has
been licensed to other companies. I moaned long and
hard about this. I'll try again next time we meet, as
whatever one thinks about the phrase `open source', it
is currently trendy in board rooms, so they might be
aggreable to a pitch that results in a media release
"Cisco helps open source community".

Again, you don't need this protocol to do enough
output hashing to interoperate. Linux could establish
its own hashing scheme and happily co-work with
EtherChannel switches. It's just that the port allocation
on the switch needs to be done manually. Not a big deal.


BTW, I found the Cisco people I dealt with to be as helpful
as they could be, given their position in the company.
They all realised the strategic value of free implementations
of protocols that are going to be munged into standards one
day.

For protocols for which there will be no standard, they
still act pretty much the same as other companies. EIGRP
being the obvious example. So use OSPF instead :-)

--
Glen Turner Network Specialist
Tel: (08) 8303 3936 Information Technology Services
Fax: (08) 8303 4400 The University of Adelaide 5005
Email: glen.turner@adelaide.edu.au South Australia

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:51    [W:0.056 / U:8.768 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site