Arie

Thanks for responding on the polarization question and I agree it can
enhance ECMP capability and maybe even counter or reduce effects of
polarization.

Thanks

Gyan

On Tue, May 25, 2021 at 1:26 PM Arie Vayner <[email protected]> wrote:

> The flow polarization or elephant flow issues are well known industry
> items and they apply to both equal and unequal cost multi-path approaches.
>
> The objective of this proposal is to enhance ECMP, and enable unequal cost
> multi-pathing, which is very useful when a service is offered by a
> multitude of endpoints, which may or may not have the same capacity.
> This solution has been implemented in production, and offers a real option
> for traffic load management.
>
> Tnx
> Arie
>
> On Tue, May 25, 2021 at 9:48 AM Jakob Heitz (jheitz) <jheitz=
> [email protected]> wrote:
>
>> The link bandwidth community has been implemented by Cisco and deployed by
>>
>> our customers for several years.
>>
>> Polarization of flows in multipath is a well known problem, but it hasn't
>> deterred
>>
>> people from using it.
>>
>>
>>
>> Regards,
>>
>> Jakob.
>>
>>
>>
>> *From:* BESS <[email protected]> *On Behalf Of * Gyan Mishra
>> *Sent:* Tuesday, May 25, 2021 12:24 AM
>> *To:* Satya Mohanty (satyamoh) <[email protected]>
>> *Cc:* [email protected]
>> *Subject:* Re: [bess] Request discussion on Cumulative Link Bandwidth
>> Draft
>>
>>
>>
>>
>>
>> Hi Satya
>>
>>
>>
>> I read the draft and have a few questions.
>>
>>
>>
>> IPv4 does not support per flow per packet load balancing as all packets
>> belonging to the same flow must hash to the same path to prevent out of
>> order packets and thus is subject to polarization of flows as high
>> bandwidth flows may hash to the same path and low bandwidth flows as well
>> to the same path resulting in very uneven load balancing.  Do to this issue
>> it does not make either iBGP or eBGP can really benefit from link bandwidth
>> extended community weight based load sharing.
>>
>>
>>
>> IPV6 flow label RFC 6437 stateless locally significant 5-tuple header
>> hash generated 20 byte key input to hash function results in uniform 50/50
>> load balancing over EGP or IGP ECMP paths.
>>
>>
>>
>> I think it maybe a good idea to reference the IPv4 polarization issue
>> with flow based load balancing and that only with IPv6 flow label can true
>> 50/50 uniform load balancing be achieved.
>>
>>
>>
>> I noticed that the normative draft referenced was adopted but has not
>> progressed.
>>
>>
>>
>> https://datatracker.ietf.org/doc/draft-ietf-idr-link-bandwidth/
>>
>>
>>
>> Has the draft been implemented by Cisco or any other vendors ?
>>
>>
>>
>> Kind Regards
>>
>>
>>
>> Gyan
>>
>>
>>
>>
>>
>> On Sat, May 22, 2021 at 11:38 AM Satya Mohanty (satyamoh) <satyamoh=
>> [email protected]> wrote:
>>
>> Hi all,
>>
>>
>>
>> On behalf of all the authors, we request a discussion of the draft
>> https://datatracker.ietf.org/doc/html/draft-mohanty-bess-ebgp-dmz-03
>>  and subsequent WG adoption.
>>
>> This draft extends the usage of the DMZ link bandwidth to scenarios where
>> the cumulative link bandwidth needs to be advertised to a BGP speaker.
>>
>> Additionally, there is provision to send the link bandwidth extended
>> community to EBGP speakers via configurable knobs. Please refer to section
>> 3 and 4 for the use cases.
>>
>>
>>
>> This feature has multiple-vendor implementations and has been deployed by
>> several customers in their networks.
>>
>>
>>
>> Best Regards,
>>
>> --Satya
>>
>> _______________________________________________
>> BESS mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/bess
>>
>> --
>>
>> <http://www.verizon.com/>
>>
>> *Gyan Mishra*
>>
>> *Network Solutions Architect *
>>
>> *Email [email protected] <[email protected]>*
>>
>> *M 301 502-1347 <(301)%20502-1347>*
>>
>>
>> _______________________________________________
>> BESS mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/bess
>>
> --

<http://www.verizon.com/>

*Gyan Mishra*

*Network Solutions A**rchitect *

*Email [email protected] <[email protected]>*



*M 301 502-1347*
_______________________________________________
BESS mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/bess

Reply via email to