Re: [Bier] Comments on draft-ietf-bier-bier-yang

[Senthil Dhanaraj <[e-mail]>]

Hi Sandy,

Got you !
But, pls consider point 2 from my mail as well. That is,

       (2)
       Remember, BIER might define its own SPF (or) constrained SPF, which
is different from IGP SPF.
       *In that case, BIER ECMP paths (& the number of ECMP paths) will be
completely different from Unicast ECMP Paths.*

Example,
BAR=1 (means next-hop must be BIER capable, support same SD, support same
BSL ... So remove nodes which not satisfy this constraints from SPF tree)
or
BAR=X (A new multicast specific algorithm)

So considering both points - (1) & (2), i'd think its better we add a
separate "ECMP Configuration parameter" for BIER under BIER Yang. What do
you think ?

Thanks,
Senthil




On Mon, Apr 15, 2019 at 11:41 AM <[e-mail]> wrote:

> Hi Senthil,
>
>
> Sorry for not saying it clearly!
>
> I mean that we add the parameter in the BIER OSPF/ISIS YANG augment parts.
>
> Like this:
>
>   augment
> /rt:routing/rt:control-plane-protocols/rt:control-plane-protocol/ospf:ospf:
>
>    +--rw bier-ospf-cfg
>
>       +--rw mt-id          mt-id
>
>       +--rw bier-global
>
>          +--rw enable?      boolean
>
>          +--rw advertise?   boolean
>
>          +--rw receive?     boolean
>
>          +--rw ecmp-num?     boolean
>
>
> Since we need not compare/vadidate the value with unicast, seems like it
> won't make great sense to define this parameter.
>
>
> Thanks,
>
> Sandy
> 原始邮件
> *发件人：*SenthilDhanaraj <[e-mail]>
> *收件人：*张征00007940;
> *抄送人：*Xiejingrong <[e-mail]>;陈然00080434;BIER WG <
> [e-mail]>;[e-mail] <
> [e-mail]>;
> *日 期 ：*2019年04月15日 17:27
> *主 题 ：**Re: [Bier] Comments on draft-ietf-bier-bier-yang*
> Hi Sandy,
>
> (1)
> Its not appropriate to define the BIER ECMP parameter in ISIS/OSPF Yang.
> Also, i do not think we need to compare/validate the Unicast and BIER ECMP
> configuration.
>
> Case-1: Unicast ECMP Config = BIER ECMP Config
> Unicast ECMP = 4
> BIER ECMP = 4
>
> Case-2: Unicast ECMP   Config  > BIER ECMP Config
> Unicast ECMP = 4
> BIER ECMP = 2
> BIER would use the first 2 unicast ECMP paths.
>
> Case-2: Unicast ECMP   Config  < BIER ECMP Config
> Unicast ECMP = 2
> BIER ECMP = 4
> BIER would use the 2 unicast ECMP paths for 4 BIFT tables.
> Example, for a given bfr-id,
> BIFT ECMP table-1 uses Unicast ECMP path1
> BIFT ECMP table-2 uses Unicast ECMP path2
> BIFT ECMP table-3 uses Unicast ECMP path1
> BIFT ECMP table-4 uses Unicast ECMP path2
>
> (2)
> Remember, BIER might define its own SPF (or) constrained SPF, which is
> different from IGP SPF.
> In that case, BIER ECMP paths will be completely different to Unicast ECMP
> Paths.
>
> Thanks,
> Senthil
>
>
> On Mon, Apr 15, 2019 at 9:55 AM <[e-mail]> wrote:
>
>> Hi Senthil,
>>
>>
>> Thank you for your quick reply too! :-P
>>
>> About the ECMP number, it is OK. Let's add a parameter for it because you
>> insist on it. :-)
>>
>> But seems like it is better to add the parameter in the underlay (OSPF
>> and ISIS) configuration parts for comparing the value with max-ecmp easily.
>>
>> How do you think about it?
>>
>>
>> Thanks,
>>
>> Sandy
>> 原始邮件
>> *发件人：*SenthilDhanaraj <[e-mail]>
>> *收件人：*张征00007940;
>> *抄送人：*Xiejingrong <[e-mail]>;陈然00080434;BIER WG <
>> [e-mail]>;[e-mail] <
>> [e-mail]>;
>> *日 期 ：*2019年04月15日 15:11
>> *主 题 ：**Re: [Bier] Comments on draft-ietf-bier-bier-yang*
>> Hi Sandy,
>> Thanks for the quick reply !
>> Pls check my comments inline !
>>
>> -
>> Senthil
>>
>> On Mon, Apr 15, 2019 at 8:30 AM <[e-mail]> wrote:
>>
>>> Hi Senthil, Jingrong,
>>>
>>>
>>> Thank you for your review and comments.
>>>
>>> I agree with your modification comments about the address-family.
>>>
>> Senthil//  Cool :)
>>
>> About the encapsulation part:
>>>
>>>    |        +--rw encapsulation* [bitstringlength]
>>>
>>>    |           +--rw bitstringlength               uint16
>>>
>>>    |           +--rw encapsulation-type            enum
>>>
>>>    |           +--rw max-si?                       rt-type: uint16
>>>
>>>    |           +--rw bift-id-base?                 rt-types: bift-id
>>>
>>> Do you think if it's better to use both BSL and encapsulation-type as
>>> the keys?
>>>
>> Senthil// Yes. To allow a sub-domain to support many encapsulation-types,
>> both "encapsulation-type" and "bsl" should be key.
>>
>>>
>>> About the load-balance-num, IMO it's not necessary to define it in BIER.
>>>
>>> BIER ECMP depends on IGP's ECMP capability. And as Jingrong said, there
>>> may be different ECMP path number in different routers.
>>>
>>> So at most a ECMP enable capability can be showed in BIER model
>>> (sub-domain), it's not necessary to define the specific number.
>>>
>> Senthil//  Multicast routing/forwarding tables and the forwarding
>> procedure itself requires more resources and is expensive compared to
>> unicast forwarding. So i'd prefer allowing the user to configure lesser
>> ECMP number for multicast compared to unicast.
>>
>> Example, Unciast (ISIS/OSPF) ECMP=64, Multicast(BIER) ECMP=4.
>> So, even though there are 64 possible unicast ECMP paths, multicast would
>> only use the first 4 ECMP paths.
>>
>>>
>>> Thanks,
>>>
>>> Sandy
>>>
>>>
>>> 原始邮件
>>> *发件人：*SenthilDhanaraj <[e-mail]>
>>> *收件人：*Xiejingrong <[e-mail]>;
>>> *抄送人：*陈然00080434;[e-mail] <[e-mail]>;
>>> [e-mail] <[e-mail]>;
>>> *日 期 ：*2019年04月15日 13:45
>>> *主 题 ：**Re: [Bier] Comments on draft-ietf-bier-bier-yang*
>>> Dear Yang Authors,
>>> 1) I agree to Jingrong's comment that "Same sub-domain cannot be binded
>>> to both IPv4 and IPv6 underlay". Pls refer the suggested model to handle
>>> this at the end of the mail. Let me know your thoughts/comments.
>>>
>>> 2) About Jingrong's questions on "Whether same sub-domain can support
>>> different encapsulation types like MPLS and Ethernet" ?
>>> I would think - Yes, a single sub-domain can support many encapsulation
>>> types. Architecturally it is possible that, for a sub-domain, each hop can
>>> chose the encapsulation to be used based on next-hops capability. Yang
>>> model should support it. However, we can discuss and clarify this.
>>>
>>> 3) A general request to BIER WG is that, we can discuss & progress the
>>> yang work at better pace. Traditionally, yang standards progress slowly in
>>> IETF resulting in implementation with private yang models :(
>>>
>>> Suggested BIER Yang Mode (sd is binded to either ipv4 or ipv6):
>>>
>>>
>>>    +--rw bier
>>>
>>>    |  +--rw bier-global
>>>
>>>    |     +--rw default-encapsulation-type?        identityref
>>>
>>>    |     +--rw default-bitstringlength?           bsl
>>>
>>>    |     +--rw default-bfr-id?                    bfr-id
>>>
>>>    |     +--rw default-ipv4-bfr-prefix?           inet:ipv4-prefix
>>>
>>>    |     +--rw default-ipv6-bfr-prefix?           inet:ipv6-prefix
>>>
>>>    |     +--rw sub-domain* [sub-domain-id] [addr-family]
>>>
>>>    |        +--rw sub-domain-id            sub-domain-id
>>>
>>>    |        +--rw addr-family            addr-family
>>>
>>>    |        +--rw bfr-prefix?  inet:ipv4-ipv6-prefix
>>>
>>>    |        +--rw underlay-protocol-type?  underlay-protocol-type
>>>
>>>    |        +--rw mt-id?                   mt-id
>>>
>>>    |        +--rw bfr-id?                  bfr-id
>>>
>>>    |        +--rw bitstringlength?         bsl
>>>
>>>    |        +--rw igp-algorithm?           ipa
>>>
>>>    |        +--rw bier-algorithm?          Bar
>>>
>>>    |        +--rw load-balance-num         uint8
>>>
>>>    |        +--rw encapsulation* [bitstringlength]
>>>
>>>    |           +--rw bitstringlength               uint16
>>>
>>>    |           +--rw encapsulation-type            enum
>>>
>>>    |           +--rw max-si?                       rt-type: uint16
>>>
>>>    |           +--rw bift-id-base?                 rt-types: bift-id
>>>
>>> Thanks,
>>> Senthil
>>>
>>> On Mon, Apr 1, 2019 at 12:10 PM Xiejingrong <[e-mail]>
>>> wrote:
>>>
>>>> Hi Chen Ran,
>>>>
>>>>
>>>>
>>>> [Ran] "load-balance-number"?Do you means the maximum number of ECMP
>>>> paths? OSPF YANG data model has defined it .In my opinion, it is
>>>> neccesarry   to define this item here.
>>>>
>>>>
>>>>
>>>> [XJR1]:
>>>>
>>>> Yes I found the load-balance(max-ecmp) configuration in OSPF-yang and
>>>> ISIS-yang, but I think they are different things, and there should be a
>>>> load-balance-number  for BIER specifically:
>>>>
>>>> (1)     A BFR may not support BIER ECMP forwarding, while unicast ECMP
>>>> is supported.
>>>>
>>>> (2)     There may be different number of paths to different BFERs, for
>>>> example BFER2/BFER2 may have 3/5 paths separately on a BFR, and this BFR
>>>>  may want a special load-balance-number 15 for better balancing.
>>>>
>>>>
>>>>
>>>> [XJR2]:
>>>>
>>>> Second question:
>>>>
>>>> Is it allowed for both IPv4-encapsulation and IPv6-encapsulation being
>>>> under a single Sub-domain ?
>>>>
>>>>
>>>>
>>>> augment /rt:routing:
>>>>
>>>>    +--rw bier
>>>>
>>>>    |  +--rw bier-global
>>>>
>>>>    |     +--rw sub-domain* [sub-domain-id]
>>>>
>>>>    |        +--rw sub-domain-id            sub-domain-id
>>>>
>>>>    |        +--rw underlay-protocol-type?  underlay-protocol-type
>>>>
>>>>    |        +--rw mt-id?                    mt-id
>>>>
>>>>    |        +--rw bfr-id?                   bfr-id
>>>>
>>>>    |        +--rw bitstringlength?          bsl
>>>>
>>>>    |        +--rw igp-algorithm?            ipa
>>>>
>>>>    |        +--rw bier-algorithm?           bar
>>>>
>>>>    |        +--rw af
>>>>
>>>>    |           +--rw ipv4* [bitstringlength bier-mpls-label-base]
>>>>
>>>>    |           |  +--rw bitstringlength               uint16
>>>>
>>>>    |           |  +--rw bier-mpls-label-base
>>>> rt-types:mpls-label
>>>>
>>>>    |           |  +--rw max-si?                       max-si
>>>>
>>>>    |           +--rw ipv6* [bitstringlength bier-mpls-label-base]
>>>>
>>>>    |              +--rw bitstrin+--glength            uint16
>>>>
>>>>    |              +--rw bier-mpls-label-base
>>>> rt-types:mpls-label
>>>>
>>>>    |              +--rw max-si?                       max-si
>>>>
>>>>    |
>>>>
>>>>
>>>>
>>>> The RFC8279 said, a BIER sub-domain must be associated with a single
>>>> routing underlay (see below). I would understand IPv4 and IPv6 as different
>>>>  underlay.
>>>>
>>>>    If multiple routing underlays are used in a single BIER domain, each
>>>>
>>>>    BIER sub-domain MUST be associated with a single routing underlay
>>>>
>>>>    (though multiple sub-domains may be associated with the same routing
>>>>
>>>>    underlay).
>>>>
>>>>
>>>>
>>>> [XJR3]:
>>>>
>>>> Third question, maybe for the BIER WG.
>>>>
>>>> It may also be helpful to discuss and conclude, if it is allowed for
>>>> both BIER-MPLS encapsulation and BIER-Ethernet encapsulation being under a
>>>> single  sub-domain?
>>>>
>>>> I feel it unnecessary since one can use different BIER Sub-domains
>>>> carrying different encapsulations, and thus an MVPN service using BIER
>>>> doesn’t  have to specify the encapsulation-type.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> *From:* [e-mail] [mailto:[e-mail]]
>>>> *Sent:* Thursday, September 06, 2018 4:04 PM
>>>> *To:* Xiejingrong <[e-mail]>
>>>> *Cc:* [e-mail]; [e-mail]
>>>> *Subject:* Re: Re: [Bier] Comments on draft-ietf-bier-bier-yang
>>>>
>>>>
>>>>
>>>> Hi jinrong，
>>>>
>>>> Thanks for your review. Please see inline...
>>>>
>>>>
>>>>
>>>> Regards.
>>>>
>>>> Ran
>>>>
>>>>
>>>>
>>>> 原始邮件
>>>>
>>>> *发件人：*Xiejingrong <[e-mail]>
>>>>
>>>> *收件人：*BIER WG <[e-mail]>
>>>>
>>>> *抄送人：*[e-mail] <
>>>> [e-mail]>
>>>>
>>>> *日 期 ：*2018年07月28日 21:01
>>>>
>>>> *主 题 ：Re: [Bier] Comments on draft-ietf-bier-bier-yang*
>>>>
>>>> _______________________________________________
>>>> BIER mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/bier
>>>>
>>>>
>>>>
>>>> some more comments:
>>>>
>>>> 1.  one sub-domain should allow miltiple {BSL and the according label
>>>> block}s as encapsulations, see the igp sub-sub-TLV.
>>>>
>>>> [Ran] We will add them ,and  will add the enternet and IPv6
>>>> encapsulation type.
>>>>
>>>> 2. should the igp-type change to underlay-protocol-type to allow bgp?
>>>>
>>>> [Ran ]will add it.
>>>>
>>>> *From:*Xiejingrong
>>>>
>>>> *To:*BIER WG,
>>>>
>>>> *Cc:*[e-mail],
>>>>
>>>> *Date:*2018-07-28 20:36:25
>>>>
>>>> *Subject:*[Bier] Comments on draft-ietf-bier-bier-yang
>>>>
>>>>
>>>>
>>>> Hi folks,
>>>>
>>>>
>>>>
>>>> I have the following comments and on draft-ietf-bier-bier-yang.
>>>>
>>>> --should the bier load-balance-number/ipa/bar be added to
>>>> rt:routing/bier-global/sub-domain (like below)? I think they are some basic
>>>> items.
>>>>
>>>>  [Ran] "load-balance-number"?Do you means the maximum number of ECMP
>>>> paths? OSPF YANG data model has defined it .In my opinion, it is neccesarry
>>>>   to define this item here.
>>>>
>>>>  For the ipa/bar will be added to  rt:routing/bier-global/sub-domain.
>>>>
>>>> augment /rt:routing:
>>>>
>>>>    +--rw bier
>>>>
>>>>    |  +--rw bier-global
>>>>
>>>>    |     +--rw encapsulation-type?   identityref
>>>>
>>>>    |     +--rw bitstringlength?      bsl
>>>>
>>>>    |     +--rw bfr-id?               bfr-id
>>>>
>>>>    |     +--rw ipv4-bfr-prefix?   inet:ipv4-prefix
>>>>
>>>>    |     +--rw ipv6-bfr-prefix?   inet:ipv6-prefix
>>>>
>>>>    |     +--rw sub-domain* [sub-domain-id]
>>>>
>>>>    |        +--rw sub-domain-id      sub-domain-id
>>>>
>>>>    |        +--rw igp-type?          igp-type
>>>>
>>>>    |        +--rw mt-id?             mt-id
>>>>
>>>>    |        +--rw bfr-id?            bfr-id
>>>>
>>>>    |        +--rw bitstringlength?   bsl
>>>>
>>>>    |        +--rw multi-bift-number? load-balance-number
>>>>
>>>>    |        +--rw igp-algorithm?     ipa
>>>>
>>>>    |        +--rw bier-algorithm?    bar
>>>>
>>>>
>>>>
>>>> --should the bier-mpls-label-range-size be changed to ‘max si’ or not ?
>>>> The type is uint8 and thus seems having to change the meaning.
>>>>
>>>>  [Ran] Sure.
>>>>
>>>>
>>>>
>>>> Thanks
>>>>
>>>> Jingrong
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> BIER mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/bier
>>>>
>>>
>>>
>>
>