Mailing List Archive

ForCES, a big improvment for free routers?
Is anyone here member of the IETF ForCES Working-Group
<URL:http://www.sstanamera.com/~forces/>? This group seems very
interesting to us because it tries to standardize the protocole used
between the Control Element (Quagga, in our case, I use ForCES
vocabulary) and the Forwarding Element (the Unix kernel). At the
present time, these two elements talk with a custom protocol (Netlink
- RFC 3549 - for Linux).

So, it makes difficult to build a PC router running at line speed
(think of a Gigabit Ethernet filled with very small packets) because,
although the Control Element is fine (any PC can handle the full
routing table in BGP), the Forwarding Element is quite weak (every
packet has to move from the card to the CPU and back). Some Ethernet
cards can do the actual forwarding but, to use them on a free Unix,
you depend on a card-specific driver (with the associated trouble of
NDA).

ForCES tries to create a standard protocol, allowing to freely mix the
elements from various vendors. Quagga could be the Control Element and
talk the ForCES protocol with these cards or even with outside
specialized Forwading Elements (ForCES explicitely address the case
where Control and Forwarding Elements are not in the same box).

It seems that such a protocol could open a very wide way of
opportunities to a new generation of routers, with free software as
the Control Element.

Presently, ForCES is at the beginning: the framework, the terminology,
the requirments are almost complete but they still have to develop,
implement and deploy the actual protocol :-)

Ideas, advices?
Re: ForCES, a big improvment for free routers? [ In reply to ]
Stephane Bortzmeyer wrote:
> Is anyone here member of the IETF ForCES Working-Group
> <URL:http://www.sstanamera.com/~forces/>? This group seems very
> interesting to us because it tries to standardize the protocole used
> between the Control Element (Quagga, in our case, I use ForCES
> vocabulary) and the Forwarding Element (the Unix kernel). At the
> present time, these two elements talk with a custom protocol (Netlink
> - RFC 3549 - for Linux).
>
Hi Stephane,

A shame nobody replied :-), just kidding. This is VERY interesting from
a conceptual and practical point of view. As soon as I can spare some
time, I'll dig into this.

Thanks a lot for the link.
Gernot

--
Dipl.-Ing. Gernot W. Schmied, MS Network Architecture & Operations
Senior Strategist Research Group
mailto:gernot.schmied@nanorg.org http://www.nanorg.org
PGP Fingerprint: 5D70 5690 47DA 9A21 D07E B9EE C764 C9B7 9B64 B27E