[Gen-art] Genart last call review of draft-ietf-bess-evpn-igmp-mld-proxy-12

[Matt Joras via Datatracker <[e-mail]>]

Reviewer: Matt Joras
Review result: Ready with Nits

I am the assigned Gen-ART reviewer for this draft. The General Area
Review Team (Gen-ART) reviews all IETF documents being processed
by the IESG for the IETF Chair.  Please treat these comments just
like any other last call comments.

For more information, please see the FAQ at

<https://trac.ietf.org/trac/gen/wiki/GenArtfaq>.

Document: draft-ietf-bess-evpn-igmp-mld-proxy-??
Reviewer: Matt Joras
Review Date: 2021-08-26
IETF LC End Date: 2021-09-07
IESG Telechat date: Not scheduled for a telechat

Review

   Ethernet Virtual Private Network (EVPN) solution is becoming
   pervasive in data center (DC) applications for Network Virtualization
   Overlay (NVO) and DC interconnect (DCI) services, and in service
   provider (SP) applications for next generation virtual private LAN
   services.

This intro to the abstract could use some rewording. For example, "is becoming"
does not read well in a standards document. It is sufficient to describe what
this document is specifying. Also "next generation" is overused and doesn't
usually read well retrospectively.

   This draft describes how to support efficiently endpoints running
   IGMP for the above services over an EVPN network by incorporating
   IGMP proxy procedures on EVPN PEs.

Avoid using the term "draft" as this will have to be edited out since the idea
is for this to be standards track.

1.  Introduction

   Ethernet Virtual Private Network (EVPN) solution [RFC7432] is
   becoming pervasive in data center (DC) applications for Network
   Virtualization Overlay (NVO) and DC interconnect (DCI) services, and
   in service provider (SP) applications for next generation virtual
   private LAN services.

This is a copy of the abstract, and has similar issues. The introduction serves
a different purpose beyond the abstract, it also has the same grammatical
issues as the abstract.

   In DC applications, a point of delivery (POD) can consist of a
   collection of servers supported by several top of rack (TOR) and
   Spine switches.  This collection of servers and switches are self
   contained and may have their own control protocol for intra-POD
   communication and orchestration.  However, EVPN is used as standard
   way of inter-POD communication for both intra-DC and inter-DC.  A
   subnet can span across multiple PODs and DCs.  EVPN provides robust
   multi-tenant solution with extensive multi-homing capabilities to
   stretch a subnet (VLAN) across multiple PODs and DCs.  There can be
   many hosts/VMs(virtual machine) (several hundreds) attached to a
   subnet that is stretched across several PODs and DCs.

Why is "Spine" capitalized? I'm also wondering whether another term might be
appropriate here that doesn't imply a certain topology. Also,  "provides robust
multi-tenant solution" should probably be "provides a robust multi-tenant
solution".

   These hosts/VMs express their interests in multicast groups on a
   given subnet/VLAN by sending IGMP Membership Reports (Joins) for
   their interested multicast group(s).  Furthermore, an IGMP router
   periodically sends membership queries to find out if there are hosts
   on that subnet that are still interested in receiving multicast
   traffic for that group.  The IGMP/MLD Proxy solution described in
   this draft accomplishes three objectives:

I don't think you need "/VMs". They are a kind of host. There is also another
use of "draft" in this paragraph.

   3.  Selective Multicast: to forward multicast traffic over EVPN
       network such that it only gets forwarded to the PEs that have
       interest in the multicast group(s), multicast traffic will not be
       forwarded to the PEs that have no receivers attached to them for
       that multicast group.  This draft shows how this objective may be
       achieved when Ingress Replication is used to distribute the
       multicast traffic among the PEs.  Procedures for supporting
       selective multicast using P2MP tunnels can be found in
       [I-D.ietf-bess-evpn-bum-procedure-updates]

The first sentence is very long and could probably be reworded to be less
redundant. Also there is another instance of "draft".

   The first two objectives are achieved by using IGMP/MLD proxy on the
   PE and the third objective is achieved by setting up a multicast
   tunnel (e.g., ingress replication) only among the PEs that have
   interest in that multicast group(s) based on the trigger from IGMP/
   MLD proxy processes.  The proposed solutions for each of these
   objectives are discussed in the following sections.

The first sentence can probably be split into two. Also, is "(e.g., ingress
replication)" really an example? "multicast tunnel" probably suffices.

   o  Ethernet Segment (ES): When a customer site (device or network) is
      connected to one or more PEs via a set of Ethernet links, then
      that set of links is referred to as an 'Ethernet Segment'.

   o  Ethernet Segment Identifier (ESI): A unique non-zero identifier
      that identifies an Ethernet Segment is called an 'Ethernet Segment
      Identifier'.

Both of these terminology definitions can drop the end part where they quote
the thing they're defining. It is implied by the colon.

   o  Ethernet Tag: An Ethernet tag identifies a particular broadcast
      domain, e.g., a VLAN.  An EVPN instance consists of one or more
      broadcast domains.

Same issue here more or less. You don't need to start out a sentence saying
"Ethernet tag" again.

4.1.  Proxy Reporting

   When IGMP protocol is used between hosts/VMs and their first hop EVPN
   router (EVPN PE), Proxy-reporting is used by the EVPN PE to summarize
   (when possible) reports received from downstream hosts and propagate
   them in BGP to other PEs that are interested in the information.
   This is done by terminating the IGMP Reports in the first hop PE, and
   translating and exchanging the relevant information among EVPN BGP
   speakers.  The information is again translated back to IGMP message
   at the recipient EVPN speaker.  Thus it helps create an IGMP overlay
   subnet using BGP.  In order to facilitate such an overlay, this
   document also defines a new EVPN route type NLRI, the EVPN Selective
   Multicast Ethernet Tag route, along with its procedures to help
   exchange and register IGMP multicast groups Section 9.

Another usage of hosts/VMs.

   1.  When the first hop PE receives several IGMP Membership Reports
       (Joins), belonging to the same IGMP version, from different
       attached hosts/VMs for the same (*,G) or (S,G), it SHOULD send a
       single BGP message corresponding to the very first IGMP
       Membership Request (BGP update as soon as possible) for that
       (*,G) or (S,G).  This is because BGP is a stateful protocol and
       no further transmission of the same report is needed.  If the
       IGMP Membership Request is for (*,G), then multicast group
       address MUST be sent along with the corresponding version flag
       (v2 or v3) set.  In case of IGMPv3, the exclude flag MUST also be
       set to indicate that no source IP address must be excluded
       (include all sources"*").  If the IGMP Join is for (S,G), then
       besides setting multicast group address along with the version
       flag v3, the source IP address and the IE flag MUST be set.  It
       should be noted that when advertising the EVPN route for (S,G),
       the only valid version flag is v3 (v2 flags MUST be set to zero).

Another hosts/VMs usage. Also missing a space after "include all sources".

   7.  Upon receiving EVPN SMET route(s) and before generating the
       corresponding IGMP Membership Request(s), the PE checks to see
       whether it has any CE multicast router for that BD on any of its
       ES's . The PE provides such a check by listening for PIM Hello
       messages on that AC (i.e, ES,BD).  If the PE does have the
       router's ACs, then the generated IGMP Membership Request(s) are
       sent to those ACs.  If it doesn't have any of the router's AC,
       then no IGMP Membership Request(s) needs to be generated.  This
       is because sending IGMP Membership Requests to other hosts can
       result in unintentionally preventing a host from joining a
       specific multicast group using IGMPv2 - i.e., if the PE does not
       receive a join from the host it will not forward multicast data
       to it.  Per [RFC4541] , when an IGMPv2 host receives a Membership
       Report for a group address that it intends to join, the host will
       suppress its own membership report for the same group, and if the
       PE does not receive an IGMP Join from host it will not forward
       multicast data to it.  In other words, an IGMPv2 Join MUST NOT be
       sent on an AC that does not lead to a CE multicast router.  This
       message suppression is a requirement for IGMPv2 hosts.  This is
       not a problem for hosts running IGMPv3 because there is no
       suppression of IGMP Membership Reports.
Need a "the" before "host in "and if the PE does not receive an IGMP Join from
host it will not forward multicast data to it."

   2.  When a PE receives an EVPN SMET route for a given (*,G), it
       compares the received version flags from the route with its per-
       PE stored version flags.  If the PE finds that a version flag
       associated with the (*,G) for the remote PE is reset, then the PE
       MUST generate IGMP Leave for that (*,G) toward its local
       interface (if any) attached to the multicast router for that
       multicast group.  It should be noted that the received EVPN route
       SHOULD at least have one version flag set.  If all version flags
       are reset, it is an error because the PE should have received an
       EVPN route withdraw for the last version flag.  Error MUST be
       considered as BGP error and the PE MUST apply the "treat-as-
       withdraw" procedure of [RFC7606].

Consider rewording the latter part of this paragraph, note that "Error MUST be
considered" is not quite grammatical.. Also if this is an error condition,
should the "SHOULD at least have one version flag set" be a MUST?

5.  Operation
...
   o  only hosts/VMs

   o  mix of hosts/VMs and multicast source

   o  mix of hosts/VMs, multicast source, and multicast router

More hosts/VMs. I will stop mentioning this nit.

6.  All-Active Multi-Homing

   Because the LAG flow hashing algorithm used by the CE is unknown at
   the PE, in an All-Active redundancy mode it must be assumed that the
   CE can send a given IGMP message to any one of the multi-homed PEs,
   either DF or non-DF; i.e., different IGMP Membership Request messages
   can arrive at different PEs in the redundancy group and furthermore
   their corresponding Leave messages can arrive at PEs that are
   different from the ones that received the Join messages.  Therefore,
   all PEs attached to a given ES must coordinate IGMP Membership
   Request and Leave Group (x,G) state, where x may be either '*' or a
   particular source S, for each BD on that ES.  This allows the DF for
   that (ES,BD) to correctly advertise or withdraw a Selective Multicast
   Ethernet Tag (SMET) route for that (x,G) group in that BD when
   needed.  All-Active multihoming PEs for a given ES MUST support IGMP
   synchronization procedures described in this section if they need to
   perform IGMP proxy for hosts connected to that ES.

"LAG" is undefined. Should "All-Active" really be capitalized?

6.2.2.  Common Leave Group Synchronization
...
   When the Maximum Response Timer expires a PE that has advertised an
   IGMP Leave Synch route, withdraws it.  Any PE attached to the
   multihomed ES, that started the Maximum Response Time and has no
   local IGMP Membership Request (x,G) state and no installed IGMP Join
   Synch routes, it removes IGMP Membership Request (x,G) state for that
   (ES,BD).  If the DF no longer has IGMP Membership Request (x,G) state
   for that BD on any ES for which it is DF, it withdraws its SMET route
   for that (x,G) group in that BD.

The first sentence should be reworded, ending the sentence with ", withdraws
it." reads awkwardly. The next sentence is also long and is confusing to read,
I'm actually not quite sure what it is trying to convey.

6.3.  Mass Withdraw of Multicast join Sync route in case of failure

   A PE which has received an IGMP Membership Request, would have synced
   the IGMP Join by the procedure defined in section 6.1.  If a PE with
   local join state goes down or the PE to CE link goes down, it would
   lead to a mass withdraw of multicast routes.  Remote PEs (PEs where
   these routes were remote IGMP Joins) SHOULD NOT remove the state
   immediately; instead General Query SHOULD be generated to refresh the
   states.  There are several ways to detect failure at a peer, e.g.
   using IGP next hop tracking or ES route withdraw.

The first sentence has an extraneous comma after "IGMP Membership Request".

9.1.  Selective Multicast Ethernet Tag Route
...
   o  The least significant bit, bit 7 indicates support for IGMP
      version 1.  Since IGMP V1 is being deprecated , sender MUST set it
      as 0 for IGMP and receiver MUST ignore it.

Extraneous comma and space in the second sentence, and missing article before
"sender".

   o  The second least significant bit, bit 6 indicates support for IGMP
      version 2.

   o  The third least significant bit, bit 5 indicates support for IGMP
      version 3.

   o  The fourth least significant bit, bit 4 indicates whether the
      (S,G) information carried within the route-type is of an Include
      Group type (bit value 0) or an Exclude Group type (bit value 1).
      The Exclude Group type bit MUST be ignored if bit 5 is not set.

Is it typical to express this in terms of which least significant bit and the
bit number? It reads a bit oddly. It's also only done in some of the
descriptions so it is not consistent.

   o  If route is used for IPv6 (MLD) then bit 7 indicates support for
      MLD version 1.  The second least significant bit, bit 6 indicates
      support for MLD version 2.  Since there is no MLD version 3, in
      case of IPv6 route third least significant bit MUST be 0.  In case
      of IPv6 routes, the fourth least significant bit MUST be ignored
      if bit 6 is not set.

Will there never be a MLD version 3? Also again missing an article before
"third least significant bit", though I have similar commentary as above about
how to refer to bits individually.

9.1.1.  Constructing the Selective Multicast Ethernet Tag route
...
   Reserved bits MUST be set to 0.  They can be defined in future by

Why are these a MUST whereas the earlier reserved bits in section 9.1 were
SHOULDs?

9.1.2.  Default Selective Multicast Route
...
   Consider the EVPN network of Figure-2, where there is an EVPN
   instance configured across the PEs.  Lets consider PE2 is connected
   to multicast router R1 and there is a network running PIM ASM behind
   R1.  If there are receivers behind the PIM ASM network, the PIM Join
   would be forwarded to the PIM RP (Rendezvous Point).  If receivers
   behind PIM ASM network are interested in a multicast flow originated
   by multicast source S2 (behind PE1), it is necessary for PE2 to
   receive multicast traffic.  In this case PE2 MUST originate a (*,*)
   SMET route to receive all of the multicast traffic in the EVPN
   domain.  To generate Wildcards (*,*) routes, prcedure from [RFC6625]
   SHOULD be used.

"Lets" should be "Let's", also probably should be "consider that PE2 is
connected". The comma in " If there are receivers behind the PIM ASM network, "
is extraneous. The last sentence has a typo in "procedure" and is missing "the"
before it.

9.2.  Multicast Join Synch Route
...
Similar commentary for this section about how bits are referred to, both by
index and which "least significant bit" they are.

   o  Reserved bits MUST be set to 0.  They can be defined in future by
      other document.
Probably don't need the second sentence at all as that's implicit, also "future
document" is more grammatical.

9.2.1.  Constructing the Multicast Join Synch Route
...

   The Flags field indicates the version of IGMP protocol from which the
   Membership Report was received.  It also indicates whether the
   multicast group had INCLUDE or EXCLUDE bit set.

Earlier in the section "INCLUDE" and "EXCLUDE" were not capitalized. One should
be picked.

   Reserved bits MUST be set to 0.  They can be defined in future by
   other document.

Same commentary as before.