Re: WGLC for draft-ietf-bfd-large-packets (Jeffrey Haas)

"Albert Fu (BLOOMBERG/ 120 PARK)" <afu14@bloomberg.net> Wed, 22 May 2024 03:58 UTC

Date: Wed, 22 May 2024 03:58:26 -0000
From: "Albert Fu (BLOOMBERG/ 120 PARK)" <afu14@bloomberg.net>
To: rtg-bfd@ietf.org
MIME-Version: 1.0
In-Reply-To: <171633651079.33237.16920075543676624389@ietfa.amsl.com>
Message-ID: <664D6D6200010F9A0C340001@message.bloomberg.net>
Subject: Re: WGLC for draft-ietf-bfd-large-packets (Jeffrey Haas)
Content-Type: multipart/alternative; boundary="BOUNDARY_664D6D6200010F9A0C340001_0_25608_omsg-pw-721"
Content-ID: <ID_664D6D6200010F9A0C340001_0_16867@omsg-pw-721>
Message-ID-Hash: S56MUUYORCSKZCBATLN6WIPQQJLVKJZ7
Precedence: list
Reply-To: Albert Fu <afu14@bloomberg.net>

Hi Les,


> On May 21, 2024, at 2:40 AM, Les Ginsberg (ginsberg) 
> 
> Sooo…this was a real “blast-from-the-past” for me.
> Over four years went by with no public updates – and in looking at the diffs 
between the latest version and V2 (which is where the discussion ended for me) 
it seems that not much has changed (albeit YANG section was introduced).
>  
> I went back and reread the emails from years ago. It seems my concerns at the 

> time were addressed – largely by Section 4.


Thanks for noting the delay - we wanted to wait till we have a vendor implementation before requesting for adoption. 


Juniper has indeed implemented this feature starting in JUNOS 23.2:


https://www.juniper.net/documentation/us/en/software/junos/release-notes/23.2/junos-release-notes-23.2r1/topics/new-features/feature-descriptions/high-availability-13.html


Example of config to test padded size of 1512 bytes 

= 1500 IP_Payload + 3*MPLS_Headers = 1512

= 1484 (this is Juniper BFD pdu-size) + 28 (IP+UDP Hdrs)


set protocols ospf area 0.0.0.0 interface et-0/0/4.0 bfd-liveness-detection minimum-interval 150

set protocols ospf area 0.0.0.0 interface et-0/0/4.0 bfd-liveness-detection multiplier 3

set protocols ospf area 0.0.0.0 interface et-0/0/4.0 bfd-liveness-detection pdu-size 1484

set protocols ospf area 0.0.0.0 interface et-0/0/4.0 strict-bfd


We are still seeing "MTU" issue from time to time in our network (e.g. just a few weeks ago, we experienced this issue with one of our international circuits in India). 


We plan to deploy this feature on our P2P OSPF and eBGP links, which are predominantly where we enable BFD for fast failure detection. 


Currently, we use periodic pings to detect and alarm the "mtu" issue, which is not ideal. With the BFD Large Packet feature, we will be able to instantly detect the issue and divert traffic to working circuits.


Thanks


Albert Fu

Bloomberg


PS: Later JUNOS version has better scaling property.

Re: WGLC for draft-ietf-bfd-large-packets (Jeffre… Albert Fu (BLOOMBERG/ 120 PARK)