Hi Parag,

Inline.

Cheers,
Pablo.

From: spring <[email protected]> on behalf of Parag Kaneriya 
<[email protected]>
Date: Friday, 26 July 2019 at 10:58
To: SPRING WG <[email protected]>
Subject: [spring] draft-filsfils-spring-net-pgm-extension-srv6-usid-00

Hello,

I have gone though this draft and below are my thought with respect to CRH 
comparison given in the draft


1)      At ingress node has to more work by gathering individual micro sid . 
for example  Node A configured with micro sid  FC00::A000/32 Node B with 
FC00::B0000/32.

Now at ingress node has to concinnated these 2 micro sid into one  
FC00::A000:B000: followed by of carrier  which is 0000 .



So this is more complex stuff introduced at ingress while programing path.

PC: False. This is control plane work done once.



2)      Even shifting of bit are hardware friendly ,  it is still more work 
then normal CRH processing  where additional CPU cycle to form new Destination 
by playing with bits and from new destination header.
PC: False. Are you kidding?
uSID- One IPv6 lookup (ingress), one shift (SID processing), one IPv6 lookup 
(egress) is much less than:
CRH- One IPv6 lookup (ingress), one CRH label lookup in a new database, one 
IPv6 copy to dest, one IPv6 lookup (egress).
That’s 3 lookups. Not to mention all the destination option header parsing 
before AND after CRH.
SRH/uSID processing is already supported in many h/w platforms including 
merchant silicon. Has CRH and its related DOHs been actually implemented and 
publicly demonstrated at linerate on h/w ASICs?

Regards
Parag



Juniper Business Use Only
_______________________________________________
spring mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/spring

Reply via email to