Re: BFD echo mode

[David Ward <[e-mail]>]

Additional congestion control isn't necessary if an implementation  
takes into consideration the adaptive timer support. This is if an  
implementer believes that BFD configuration would be set by an  
operator such that the already inconsequential amount of data being  
sent actually causes congestion. BFD should not be run over TCP due  
to overhead or UDP-lite due to lack of widespread deployment.

-DWard

On Nov 30, 2008, at 1:14 PM, Vishwas Manral wrote:

> Hi Dave,
>
> Agree to what you say.
>
> I was also thinking of congestion control considerations to be put
> into BFD for the Multi-hop case. The draft does not talk about it and
> because BFD works on very short duration Hello's it may need to be TCP
> friendly in some cases.
>
> Do let me know what you think?
>
> Thanks,
> Vishwas
>
> On Thu, Nov 27, 2008 at 7:45 AM, David Ward <[e-mail]> wrote:
>> Nobo -
>>
>> On Nov 27, 2008, at 3:41 AM, Nobo Akiya (nobo) wrote:
>>
>>>
>>> Hello Vishwas.
>>>
>>> Couple of add-on comments, for single-hop.
>>>
>>>> Seeing some vendor documentation it seems they use a seperate
>>>> port for the echo mode - as the source and the destination
>>>> address.
>>>
>>> I hope vendors aren't using different udp dest port for echo  
>>> packets.
>>>
>>> http://tools.ietf.org/html/draft-ietf-bfd-v4v6-1hop-08 Sec4.1.
>>>
>>> BFD Echo packets MUST be transmitted in UDP packets with destination
>>> UDP port 3785 in an IPv4 packet.
>>>
>>>> Can the asusmption that the echo mode packet is a BFD packet
>>>> be made (so the BFD source / dest packet identifiers can be
>>>> used)? The base BFD spec seems to state the content of the
>>>> packet need not be specified as the packet is just looped back.
>>>
>>> BFD echo packets can be identified by UDP dest port 3785.
>>> Data beyond UDP header is not specified, and that's left up to  
>>> vendors.
>>>
>>>> So I can see how we could look an IP packet in the case of
>>>> Single Hop (by having the MAC Destination Address of the
>>>> peer) but having IP address and destination the same (as of
>>>> the source).
>>>
>>> I don't think there's anything that specifies that dest & src  
>>> addresses
>>> must be the same. For dest addresses tho, it maybe easier to
>>> demultiplex if dest address is the outgoing interface.
>>
>>
>> DW: there is nothing in the spec that says the addrs have to be  
>> the same.
>>
>>
>> -DWard
>>
>>>
>>> Thanx,
>>> Nobo
>>>
>>
>>