Hi Jorge,
Please see my comments with [Yubao3].
Thanks
Yubao
原始邮件
发件人:Rabadan,Jorge(Nokia-US/MountainView)
收件人:王玉保10045807;
抄送人:bess@ietf.org;
日 期 :2021年11月12日 03:43
主 题 :Re: Re:Questions on draft-sajassi-bess-evpn-ip-aliasing-03
Hi Yubao,
Please see my comments with [jorge2].
Thx
Jorge
From: wang.yub...@zte.com.cn <wang.yub...@zte.com.cn>
Date: Wednesday, November 10, 2021 at 3:55 AM
To: Rabadan, Jorge (Nokia - US/Mountain View) <jorge.raba...@nokia.com>
Cc: bess@ietf.org <bess@ietf.org>
Subject: Re:Questions on draft-sajassi-bess-evpn-ip-aliasing-03
Hi Jorge,
please see in-line with [Yubao2]
原始邮件
发件人:Rabadan,Jorge(Nokia-US/MountainView)
收件人:王玉保10045807;
抄送人:bess@ietf.org;
日 期 :2021年11月09日 20:23
主 题 :Re: Questions on draft-sajassi-bess-evpn-ip-aliasing-03
Hi Yubao,
Please see in-line.
Thanks.
Jorge
From: wang.yub...@zte.com.cn <wang.yub...@zte.com.cn>
Date: Tuesday, November 9, 2021 at 9:06 AM
To: Rabadan, Jorge (Nokia - US/Mountain View) <jorge.raba...@nokia.com>
Cc: bess@ietf.org <bess@ietf.org>
Subject: Questions on draft-sajassi-bess-evpn-ip-aliasing-03
Hi Jorge,
I read the draft, and have the following questions:
1) on section 1.2 Inter-subnet Forwarding for Prefix Routes in the
Interface-less IP-VRF-to-IP-VRF Model
The RT-1 per EVI route of ESI1 in Figure 2 is not an IP A-D per EVI route,
but a normal Ethernet A-D per EVI route,
in other words, its MPLS label identifies a BD, not an IP-VRF.
is my understanding correct?
[jorge] not really, it is an IP A-D per EVI route as explained in section 3.
In RFC9136 Interface-less IP-VRF-to-IP-VRF Model,
the inter-subnet-forwarding from H3 to H1 will pass through PE1/PE2's
IP-VRF instance via the MPLS label of the IP-VRF's instance,
but in the RFC9316 Bump-in-the-wire instance,
the inter-subnet-forwarding from H3 to H1 will not pass through PE1/PE2's
IP-VRF instance because of the RT-1 per EVI route's MPLS Label's L2 context.
But this section refers to both the above two use cases of RFC9136,
So which behavior will be followed by this use case?
[jorge] see above, the IP A-D per EVI route includes the route-target and label
of the IP-VRF as explained in section 3.
[Yubao2] I get your meanings now, I mistake the RT-1 per EVI route of use case
b (section 1.2) as a Ethernet A-D per EVI route because of the following text:
In the Interface-less IP-VRF-to-IP-VRF model described in
[I-D.ietf-bess-evpn-prefix-advertisement] there is no Overlay Index
and hence no recursive resolution of the IP Prefix route to either a
MAC/IP Advertisement or an Ethernet A-D per ES/EVI route, which means
that the fast convergence and aliasing/backup path functions are
disabled. Although the use-case is different, in a sense the
recursive resolution of an IP Prefix route to an Ethernet A-D per ES/
EVI route is already described in section 4.3 of
[I-D.ietf-bess-evpn-prefix-advertisement], Bump-in-the-Wire Use-Case,
but that section does not describe aliasing.
[jorge2] note that the above text is describing the use-cases. The concept of
the IP A-D per ES/EVI route is defined later in section 3.
[Yubao3] We should note that RFC9136 section 3 is just the route resolution
procedures,
But the use case b of draft-sajassi-bess-evpn-ip-aliasing-03 is a use
case including route advertisement, route resolution, data packet forwarding.
If they are only the same in the route resolution procedures,
it will be confusing to treat it as a Bump-in-the-wire Use-Case plus
ip-aliasing.
I still note that the route advertise ment is not clearly discribed in
.
for example, how can we know which ESI should be carried in a RT-5
route?
This is not described in RFC9136 Bump-in-the-wire use case either,
because there is no IP-VRF on NVEs.
So the route advertisement behavior of
draft-sajassi-bess-evpn-ip-aliasing-03 use case b is not clearly defined both
in this draft and in RFC9136 Bump-in-the-wire.
[Yubao2] These two concepts Ethernet A-D per ES/EVI route and IP A-D per
ES/EVI route in this draft may be confusing.
To me an ethernet A-D per EVI route means that its route-target
identifies a MAC-VRF as per [RFC7432].
an IP A-D per EVI route means that its route-target identifies
a IP-VRF as per this draft.
When they will be used by a RT-5 route with an ESI overlay
index,
they will be different in route advertisement, recursive
resolution and forwarding behavior too.
But in the above text of
draft-sajassi-bess-evpn-ip-aliasing-03, it seems that the route advertisement,
recursive resolution and forwarding
behavior in the draft is just the same as RFC9136 secion 4.3
Bump-in-the-wire use case.
I think it will be confusing. The most obvious difference
between these two use case is that,
The inter-subnet forwarding from H3 to H1 will use a MPLS
label which identifies a MAC-VRF according to Bump-in-the-wire use case,
But in this draft it is not the same behavior. This difference
will lead to other differences on route advertisement and recursive resolution
too.
[jorge] we can clarify further, but an IP A-D per EVI route uses a label that
identifies the IP-VRF, which is what is required in the symmetric IRB and
interface-less scenarios that this draft builds upon. Section 3 explains how
the IP A-D per EVI route is built.
[Yubao3] Now we have the common understanding about IP A-D per EVI route.
and the RT-1 per EVI route used by Bump-in-the-wire use case is the
old Ethernet A-D per EVI route as per [RFC7432].
2) On section 5.3 Constructing the EVPN IP Routes
Is the RT-5 construction of the second use case (section 1.2) the same as
the third use case (section 1.3) ?
I mainly concerns the Route Targets and the Ethernet Tag ID of the RT-5
routes.
especailly when the BD (to which the ESI of section 1.2 is attched) is of
VLAN-aware service interface.
[jorge] the IP Prefix routes and MAC/IP advertisement routes are constructed as
per section 5.3, hence the IP Prefix routes ethernet tag id is 0. This document
does not change the use of the Ethernet Tag ID.
[Yubao2] As I explained above, if they are the same,
is the RT-5 construction of the use case c (section 1.3) the same as
RFC9136 Bump-in-the-wire use case?
[jorge2] case 1 and case 2 build upon the IP-VRF-to-IP-VRF models in RFC9136 as
described in the use-cases.
[Yubao3] case 1 is section 1.2 of draft-sajassi-bess-evpn-ip-aliasing-03.
and case 2 is section 1.3 of draft-sajassi-bess-evpn-ip-aliasing-03.
is it correct?
3) On section 5.3.1 Route Resolution
Is the Route Resolution of the second use case (section 1.2) the same as
the third use case (section 1.3) ?
Will the route resolution of the second use case(section 1.2) need a BD and
an IRB interface on PE3?
I note that in RFC9136 section 4.3 Bump-in-the-wire use case,
the RT-1 per EVI route is advertised in a normal BD. It says that:
(1) Assuming TS2 is the active TS in ESI23, NVE2 advertises the
following BGP routes:
* Route type 1 (Ethernet A-D route for BD-10) containing: ESI =
ESI23 and the corresponding tunnel information (VNI field),
as well as the BGP Encapsulation Extended Community as per
[RFC8365].
* Route type 5 (IP Prefix route) containing: IPL = 24, IP =
SN1, ESI = ESI23, and GW IP address = 0. The EVPN Router's
MAC Extended Community defined in [RFC9135] is added and
carries the MAC address (M2) associated with the TS behind
which SN1 sits. M2 may be learned by policy; however, the
MAC in the Extended Community is preferred if sent with the
route.
This RT-1 per EVI route will not just be used by the RT-5 routes for IP
forwarding,
it will also be used by the MAC forwarding of BD-10.
When it is used in IP forwarding and MAC forwarding, it will be the same
route.
If this is correct, it will need a BD on PE3 to be resolved to.
[jorge] the resolution is the same for the three cases, based on section 5.3.1.
It happens in the context of the IP-VRF, but now considering the IP A-D routes
(which carry the IP-VRF route-target). For use-cases 2 and 3, this is
applicable to the interface-less and even interface-ful unnumbered
IP-VRF-to-IP-VRF model (we can clarify this in future versions).
[Yubao2] In RFC9136 Interface-ful unnumbered IP-VRF-to-IP-VRF model,
The MPLS payload is encapsulated as ethernet packet:
[jorge2] I don’t understand what your question is. RFC9136 may use ethernet or
IP NVO.
[Yubao3] RFC9136's Interface-less mode may use ethernet or IP NVO.
But the MAC-VRF of RFC7432 can use ethernet NVO only,
the SBD of interface-ful mode of RFC 9136 will also belong to a
MAC-VRF of RFC7432,
so it can't use IP NVO.
(3) When DGW1 receives a packet from the WAN with destination IPx,
where IPx belongs to SN1/24:
* A destination IP lookup is performed on the DGW1 IP-VRF
table. The lookup yields SN1/24, which is associated with
the Overlay Index M1. The forwarding information is derived
from the RT-2 received for M1.
* The IP packet destined to IPx is encapsulated with: inner
source MAC = M3, inner destination MAC = M1, outer source IP
(source VTEP) = DGW1 IP, and outer destination IP
(destination VTEP) = NVE1 IP.
[Yubao2] An IP A-D per EVI route whose MPLS label identifies an IP-VRF can't be
use in such encapsulation.
[jorge] I’m afraid that is not correct. It can be used, in the same way that an
RT5 whose label identifies an IP-VRF can use an ethernet or IP NVO encap. Check
out section 3.1 please.
[Yubao3] I think when we refer to a concept of RFC9136, we should let it be in
consistence with what it is defined in RFC9136.
I agree with you in that a RT5 whose label identifies an IP-VRF can
use ethernet NVO,
but when we talk about interface-ful unumbered mode or
bump-in-the-wire, we are talking about a integrated route advertisement
behavior, route resolution behavior and data encapsulation behavior,
these integration of behaviors are discribed in RFC9136 section 4.4.3
(not section 3), where there isn't an RT5 whose label identifies an IP-VRF.
if the concept will be used not as how it is used in RFC9136, It will
be confusing.
I try to understand their differences as the following, can you tell
me whether it is consistent with draft-sajassi-bess-evpn-ip-aliasing-03?
The Ethernet A-D per EVI route: Route-Target = MAC-VRF, as per
[RFC7432], [RFC9136] and [EVPN inter-subnet forwarding]
The IP A-D per EVI route: Route-Target = IP-VRF, as per
draft-sajassi-bess-evpn-ip-aliasing-03
And here are two variations which are not described in
draft-sajassi-bess-evpn-ip-aliasing-03:
The Ethernet A-D per EVI route with an IP-VRF label: it is
advertised in the context of a MAC-VRF but carries an IP-VRF label
The IP A-D per EVI route with a MAC-VRF label: it is advertised in the
context of an IP-VRF but carries a MAC-VRF label
If you want to use a SBD with a unnumbered IRB interface,
The Ethernet A-D per EVI route with an IP-VRF label may help,
But the recursive route resolution for ESI overlay index will be done in the
context of the SBD,
which is not the same as your current design, but like the distributed
bump-in-the-wire use case,
but this will have little relations with the RFC9136 Interface-ful unnumbered
IP-VRF-to-IP-VRF model,
becaus of their different MPLS payload formats.
Although this is not consistent with the route resolution design of
draft-sajassi-bess-evpn-ip-aliasing-03,
this solution may be easier to be deployed than an ethernet-format MPLS-payload
solution or the IP A-D per EVI route with a MAC-VRF label solution .
[jorge2] the IP Aliasing solution supports ethernet and ip NVO tunnels. At the
moment use-case 2 and 3 are based on the Interface-less model as discussed. If
there is interest we can add the Interface-ful unnumbered model, given that in
that model the RT5 can carry a non-zero ESI and a MAC ext community, and yet
the ESI is selected as Overlay index for the recursive resolution.
I think it will be better if the difference between new IP A-D per EVI route
and old Ethernet A-D per EVI route can be cearlified in this draft.
[jorge2] as discussed, the IP A-D per EVI route is described in section 3. Can
you go through section 3 and let me know what needs to be clarified please?
[Yubao3] I don't think that draft-sajassi-bess-evpn-ip-aliasing-03 section 3 is
not clear,
I am confused because you have refered (in section 1.2) to some use
cases of RFC9136, which has different route advertisement behavior of that
section.
The basic difference between IP A-D per EVI route and old Ethernet A-D
per EVI route is their route-targets.
But the use cases you have refered to are different from that section
in their route-target assignment too.
Maybe you just want to refer to the route rosolution procedures, not
the use case,
if this is correct, I think it will be better to refer to RFC9136
section 3 instead of Bump-in-the-wire or Interface-ful unnumbered mode.
Thanks,
Yubao
_______________________________________________
BESS mailing list
BESS@ietf.org
https://www.ietf.org/mailman/listinfo/bess