RIFT Z. Zhang
Internet-Draft Juniper Networks
Intended status: Standards Track P. Thubert
Expires: January 9, 2020 Cisco
July 8, 2019
Multicast Routing In Fat Trees
draft-zzhang-rift-multicast-00
Abstract
This document specifies multicast procedures with RIFT. Multicast in
RIFT is similar to Bidirectional Protocol Independent Multicast (PIM-
Bidir), with the Rendezvous Point Link (RP-Link) simulated by a
spanning tree of some Top of Fabric (ToF) nodes and sub-ToF nodes.
Requirements Language
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC2119.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 9, 2020.
Copyright Notice
Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of
Zhang & Thubert Expires January 9, 2020 [Page 1]
Internet-Draft mrift July 2019
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Specifications . . . . . . . . . . . . . . . . . . . . . . . 5
3. Security Considerations . . . . . . . . . . . . . . . . . . . 5
4. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 5
5. References . . . . . . . . . . . . . . . . . . . . . . . . . 5
5.1. Normative References . . . . . . . . . . . . . . . . . . 5
5.2. Informative References . . . . . . . . . . . . . . . . . 6
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 6
1. Introduction
Because of the simple north-south regular topology in Fat Tree
networks, the PIM-Bidir [RFC5015] solution is extended for multicast
in RIFT (referred to as MRIFT in this document). The following is a
summary of the changes and adaptations compared to PIM-Bidir.
With PIM-Bidir, PIM joins are sent towards a Rendezvous Point
Address, which could be an address not belonging to any router. The
RPA does belong to a RP Link (RPL), which could be attached to a
single router or multiple routers (e.g. RPL is a LAN). With MRIFT,
there is no concept of RPA any more (joins are simply sent
northbound). The joins are terminated on some sub-ToF nodes and the
RPL is simulated by a spanning tree among some ToF and sub-ToF nodes.
Instead of (*,G) trees in PIM-Bidir, MRIFT uses (*,G-Prefix) trees,
where the G-Prefix could be *, G, or anything in between (e.g.,
225.1.1.0/24). For light flows, they could just follow the (*,*)
tree. For heavy flows, individual (*,G) trees could be built. For
medium flows, some (*,G-prefix) trees could be shared. All the First
Hop Routers (FHRs, connecting to sources) and the Last Hop Routers
(LHRs, connecting to receivers) of a particular (*,G) flow must agree
on whether a (*,*) or (*,G) or (*,G-prefix) tree is used for the flow
so that they all join the same tree. This is done via out of band
control outside the scope of this document.
Because of the rich connections in Fat Trees, a router has to choose
one of its many north neighbors to send join to. This is done
through hashing. The hashing algorithm should lead to several but
not too many routers choosing the same north neighbor, so that fewer
Zhang & Thubert Expires January 9, 2020 [Page 2]
Internet-Draft mrift July 2019
routers are involved in multicast traffic forwarding, yet none of
those routers are overburdened by replicating to too many downstream
neighbors.
Instead of PIM messages, RIFT's own TIEs are used. This is similar
to the concept in [draft-zzhang-pim-pds]. Specifically, RIFT Policy
Guided Prefixes (PGP) [draft-atlas-rift-pgp] are used. The TIEs are
consumed, processed at each hop and then regenerated for the next
hop.
When a join reaches a sub-ToF node, the normal join process stops.
This forms a sub-tree rooted at this sub-ToF node. Multiple sub-
trees of the same tree may be joined by a single ToF node, or they
may have to be connected by a spanning tree serving as the RPL. For
example, in the following topology, in normal situations the two sub-
tree roots for the two pods, say Spine111 and Spine121, may be joined
by ToF21, but if the ToF21-Spine121 link is down, then ToF22 may be
used, and if the ToF22-Spine111 link is also down, then Spine111 and
Spine121 will have to be joined via
Spine111-ToF21-Spine112-ToF22-Spine121.
Zhang & Thubert Expires January 9, 2020 [Page 3]
Internet-Draft mrift July 2019
. +--------+ +--------+ ^ N
. |ToF 21| |ToF 22| |
.Level 2 ++-+--+-++ ++-+--+-++ <-*-> E/W
. | | | | | | | | |
. P111/2| |P121 | | | | S v
. ^ ^ ^ ^ | | | |
. | | | | | | | |
. +--------------+ | +-----------+ | | | +---------------+
. | | | | | | | |
. South +-----------------------------+ | | ^
. | | | | | | | All TIEs
. 0/0 0/0 0/0 +-----------------------------+ |
. v v v | | | | |
. | | +-+ +<-0/0----------+ | |
. | | | | | | | |
.+-+----++ optional +-+----++ ++----+-+ ++-----++
.| | E/W link | | | | | |
.|Spin111+----------+Spin112| |Spin121| |Spin122|
.+-+---+-+ ++----+-+ +-+---+-+ ++---+--+
. | | | South | | | |
. | +---0/0--->-----+ 0/0 | +----------------+ |
. 0/0 | | | | | | |
. | +---<-0/0-----+ | v | +--------------+ | |
. v | | | | | | |
.+-+---+-+ +--+--+-+ +-+---+-+ +---+-+-+
.| | (L2L) | | | | Level 0 | |
.|Leaf111~~~~~~~~~~~~Leaf112| |Leaf121| |Leaf122|
.+-+-----+ +-+---+-+ +--+--+-+ +-+-----+
. + + \ / + +
. Prefix111 Prefix112 \ / Prefix121 Prefix122
. multi-homed
. Prefix
.+---------- Pod 1 ---------+ +---------- Pod 2 ---------+
The following algorithm is used to form the spanning tree.
1. Each sub-tree root (a sub-ToF node) hashes to a ToF neighbor as
its parent and advertises the parent's SystemID in a N-TIE for
the tree. This allows different trees to have different RPLs for
load-balancing. In the above example, Suppose Spine111
advertises its choice of ToF21, and Spine121 advertises its
choice of ToF22.
2. Each ToF node advertises the highest SystemID in its S-TIE for a
tree, of all the ToF nodes chosen and advertised by sub-ToF nodes
for the same tree. The S-TIE also includes the SystemID of the
sub-ToFs who made the choice. A ToF node knows the choices
Zhang & Thubert Expires January 9, 2020 [Page 4]
Internet-Draft mrift July 2019
either because it is the neighbor of a sub-ToF who made a choice
(e.g. ToF21 knows Spine121's choice is ToF22 because of
Spine121's N-TIE), or because it received another ToF's S-TIE
reflected by a common south neighbor (e.g. if the ToF21-Spine121
link is down, ToF21 still knows ToF22 was chosen by Spine121
because of ToF22's S-TIE for the tree reflected by Spine122).
3. If a sub-ToF node sees ToF nodes with higher SystemIDs (than that
of its own chosen parent) advertised for the tree, it reparents
to the one that is its neighbor and has the highest SystemID, and
re-advertises the new parent. In the above example, Spine111
will reparent to ToF22, assuming ToF22 has higher SystemID than
ToF21.
4. A ToF parent (with remaining sub-ToF children who could not
reparent) joins towards the ToF parent with the highest SystemID
(as determined in step #2) via a south neighbor by including in
its S-TIE for the tree the identity of the south neighbor, who
either advertised its choice of the highest SystemID ToF parent,
or reflected a ToF node's S-TIE about sub-ToF node's choice of
the highest SystemID ToF parent. In the above example, if the
ToF22-Spine111 link is down, ToF21 will join ToF22 either via
Spine112 or Spine122.
The above procedures may repeat multiple times before the spanning
tree is settled; unless the connections among ToF and sub-ToF nodes
are badly broken, the process should be fairly simple.
2. Specifications
More details will be specified in future revisions.
3. Security Considerations
To be provided.
4. Acknowledgements
The authors thank Bruno Rijsman and Antoni Przygenda for their review
and suggestions.
5. References
5.1. Normative References
[I-D.ietf-rift-rift]
Team, T., "RIFT: Routing in Fat Trees", draft-ietf-rift-
rift-06 (work in progress), June 2019.
Zhang & Thubert Expires January 9, 2020 [Page 5]
Internet-Draft mrift July 2019
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
5.2. Informative References
[I-D.zzhang-pim-pds]
Zhang, J. and K. Patel, "Protocol Dependent Multicast
Signaling", draft-zzhang-pim-pds-00 (work in progress),
October 2015.
[RFC5015] Handley, M., Kouvelas, I., Speakman, T., and L. Vicisano,
"Bidirectional Protocol Independent Multicast (BIDIR-
PIM)", RFC 5015, DOI 10.17487/RFC5015, October 2007,
<https://www.rfc-editor.org/info/rfc5015>.
Authors' Addresses
Zhaohui Zhang
Juniper Networks
EMail: zzhang@juniper.net
Pascal Thubert
Cisco Systems, Inc
EMail: pthubert@cisco.com
Zhang & Thubert Expires January 9, 2020 [Page 6]