One document matched: draft-ietf-trill-centralized-replication-04.txt
Differences from draft-ietf-trill-centralized-replication-03.txt
TRILL Working Group W. Hao
INTERNET-DRAFT Y. Li
Intended Status: Standard Track Huawei Technologies
M. Durrani
Cisco
S. Gupta
IP Infusion
A. Qu
MediaTec
T. Han
Huawei Technologies
Expires: September 2016 March 02, 2016
Centralized Replication for BUM traffic in active-active edge
connection
draft-ietf-trill-centralized-replication-04.txt
Abstract
In TRILL active-active access scenario, RPF check failure issue may
occur when pseudo-nickname mechanism in [RFC7781] is used. This
draft describes a solution to resolve the RPF check failure issue
through centralized replication. All ingress RBridges send BUM
(Broadcast, Unknown unicast and Mutlicast) traffic to a centralized
node with unicast TRILL encapsulation. When the centralized node
receives the BUM traffic, it decapsulates the packets and forwards
them to all destination RBridges using a distribution tree
established as per TRILL base protocol[RFC6325]. To avoid RPF check
failure on a RBridge sitting between the ingress RBridge and the
centralized replication node, some change of RPF calculation
algorithm is required. RPF calculation on each RBridge should use
the centralized node as ingress RB instead of the real ingress
RBridge which is denoted as RBv in [RFC7781] to perform the
calculation.
Status of this Memo
This Internet-Draft is submitted to IETF in full conformance with
the provisions of BCP 78 and BCP 79.
Hao & Li,etc Expires July 09, 2016 [Page 1]
Internet-Draft Centralized replication for BUM traffic March 2016
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts.
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."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/1id-abstracts.html
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Copyright Notice
Copyright (c) 2016 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
(http://trustee.ietf.org/license-info) in effect on the date of
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 ................................................ 3
2. Conventions used in this document............................ 4
3. Centralized Replication Solution Overview.................... 4
4. Frame duplication from remote RB............................. 5
5. Local forwarding behavior on ingress RBridge................. 6
6. Loop prevention among RBridges in a edge group............... 7
7. Centralized replication forwarding process................... 8
Hao & Li,etc Expires September 02, 2016 [Page 2]
Internet-Draft Centralized replication for BUM traffic March 2016
8. BUM traffic load balancing among multiple centralized nodes...9
9. Co-existing with CMT solution............................... 10
10. Network Upgrade Analysis................................... 11
11. TRILL protocol extension................................... 11
11.1. "R" and "C" Flag in Nickname Flags APPsub-TLV..........11
12. Security Considerations.................................... 12
13. IANA Considerations........................................ 12
14. References ................................................ 12
14.1. Normative References.................................. 12
14.2. Informative References................................ 12
15. Acknowledgments ........................................... 13
1. Introduction
The IETF TRILL (Transparent Interconnection of Lots of Links)
[RFC6325] protocol provides loop free and per hop based multipath
data forwarding with minimum configuration. TRILL uses IS-IS
[RFC6165] [RFC6326bis] as its control plane routing protocol and
defines a TRILL specific header for user data.
Classic Ethernet (CE) devices typically are multi-homed to multiple
edge RBridges which form an edge group. All of the uplinks of CE are
bundled as a Multi-Chassis Link Aggregation (MC-LAG). An active-
active flow-based load sharing mechanism is normally implemented to
achieve better load balancing and high reliability. A CE device can
be a layer 3 end system by itself or a bridge switch through which
layer 3 end systems access to TRILL campus.
In active-active access scenario, pseudo-nickname solution in
[RFC7781] can be used to avoid MAC flip-flop on remote RBs. The
basic idea is to use a virtual RBridge RBv with a single pseudo-
nickname to represent an edge group that MC-LAG connects to. Any
member RBridge of that edge group should use this pseudo-nickname
rather than its own nickname as ingress nickname when it injects
TRILL data frames to TRILL campus. The use of the nickname solves
the address flip flop issue by making the MAC address learnt by the
remote RBridge bound to pseudo-nickname. However, it introduces
another issue of incorrect packet drop which will be described as
follows. When a pseudo-nickname is used by an edge RBridge as the
ingress nickname to forward BUM traffic, any RBridges (RBn) sitting
between the ingress RB and the distribution tree root will treat the
traffic as it is ingressed from the virtual RBridge RBv. If the same
distribution tree is used by different edge RBridges of the same RBv,
the traffic may arrive at RBn from different ports. Then the RPF
check fails, and some of the traffic receiving from unexpected ports
will be dropped by RBn.
Hao & Li,etc Expires September 02, 2016 [Page 3]
Internet-Draft Centralized replication for BUM traffic March 2016
This document proposes a centralized replication solution for
broadcast, unknown unicast and multicast(BUM) traffic forwarding to
resolve the issue of incorrect packet drop caused by RPF check
failure in virtual RBridge case. The basic idea is that all ingress
RBridges send BUM traffic to a centralized node which is recommended
to be a distribution tree root using unicast TRILL encapsulation.
When the centralized node receives the packets, it decapsulates and
forwards them to all destination RBridges using a distribution tree
established as per TRILL base protocol.
2. Conventions used in this document
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 RFC-2119
[RFC2119].The acronyms and terminology in [RFC6325] is used herein
with the following additions:
BUM - Broadcast, Unknown unicast and Multicast
CE - As in [CMT], Classic Ethernet device (end station or bridge).
The device can be either physical or virtual equipment.
3. Centralized Replication Solution Overview
When an edge RB receives BUM traffic from a CE device, it uses
unicast TRILL encapsulation instead of multicast encapsulation to
send the packets to a centralized node. The centralized node is
recommended to be a distribution tree root.
The TRILL header of the unicast TRILL encapsulation contains an
"ingress RBridge nickname" field and an "egress RBridge nickname"
field. If ingress RB receives the BUM packet from a port which is in
an MC-LAG, it should set the ingress RBridge nickname to be the
pseudo-nickname rather than its own nickname to avoid MAC flip-flop
on remote RBs as per [RFC7781]. The egress RBridge nickname is set
to the special nickname of the centralized node which is used to
differentiate the centralized replication purpose unicast TRILL
encapsulation from a normal unicast TRILL encapsulation. The special
nickname is called R-nickname.
When the centralized node receives a unicast TRILL encapsulated
packet with R-nickname as egress nickname, it decapsulates the
packet. Then the centralized node replicates and forwards the BUM
Hao & Li,etc Expires September 02, 2016 [Page 4]
Internet-Draft Centralized replication for BUM traffic March 2016
packet to all destination RBs using one of the distribution trees
established as per TRILL base protocol. It is recommended to use the
distribution tree whose tree root is the centralized node itself.
When the centralized node forwards the BUM traffic, ingress nickname
remains same as that in frame it received to ensure that the MAC
address learnt by all egress RBridges bound to pseudo-nickname.
When the replicated packet is forwarded on each RBridge along the
distribution tree starting from the centralized node, RPF check will
be performed as per [RFC6325]. For any RBridge sitting between the
ingress RBridge and the centralized replication node, the incoming
port of such BUM packet should be the centralized node facing port
as the multicast traffic always comes from the centralized node in
this solution. However the RPF port as the result of distribution
tree calculation as per [RFC6325] will be the real ingress RB facing
port as it uses virtual RBridge as the ingress RB, so RPF check will
fail. To solve this problem, some change of RPF calculation
algorithm is required. RPF calculation on each RBridge should use
the centralized node as ingress RB instead of the real ingress
virtual RBridge to perform the calculation. As a result, RPF check
will point to the centralized node facing port on the RBridge for
multi-destination traffic. It prevents the incorrect frame drop by
RPF check.
To differentiate the centralized purpose unicast TRILL encapsulation
from normal unicast TRILL encapsulation on a centralized node, R-
nickname is introduced for centralized replication. Only when the
centralized node receives unicast TRILL encapsulation traffic with
egress nickname equivalent to R-nickname, it decapsulates the packet
and then forwards the packet to all destination RBs through a
distribution tree by re-encapsulation as per aforementioned. The
centralized nodes should announce its R-nickname to all TRILL campus
through TRILL LSP extension.
4. Frame duplication from remote RB
Frame duplication may occur when a remote host sends multi-
destination frame to a local CE which has an active-active
connection to the TRILL campus. To avoid local CE receiving multiple
copies from a remote RBridge, the designated forwarder (DF)
mechanism should be supported for egress direction multicast traffic.
DF election mechanism allows only one port of one RB in an MC-LAG to
forward multicast traffic from TRILL campus to local access side for
each VLAN. The basic idea of DF is to elect one RBridge per VLAN
from an edge group to be responsible for egressing the multicast
Hao & Li,etc Expires September 02, 2016 [Page 5]
Internet-Draft Centralized replication for BUM traffic March 2016
traffic. [RFC7781] describes the detailed DF election mechanism
among member RBridges involving in a MC-LAG.
If DF election mechanism is used for frame duplication prevention,
access ports on an RB are categorized as three types: non mc-lag,
mc-lag DF port and mc-lag non-DF port. The last two types can be
called mc-lag port. For each of the mc-lag port, there is a pseudo-
nickname associated. If consistent nickname allocation per edge
group RBridges is used, it is possible that same pseudo-nickname
associated to more than one port on a single RB. A typical scenario
is that CE1 is connected to RB1 & RB2 by mc-lag1 while CE2 is
connected to RB1 & RB2 by mc-lag 2. In order to conserve the number
of pseudo-nickname used, member ports for both mc-lag1 and mc-lag2
on RB1 & RB2 are all associated to pseudo-nickname pn1.
5. Local forwarding behavior on ingress RBridge
When an ingress RBridge(RB1) receives BUM traffic from a local
active-active accessing CE(CE1) device, the traffic will be injected
to TRILL campus through TRILL encapsulation, and it will be
replicated and forwarded to all destination RBs which include
ingress RB itself along a TRILL distribution tree, the traffic will
also return to the ingress RBridge. To avoid the traffic looping
back to original sender CE, ingress nickname of pseudo-nickname can
be used for traffic filtering.
If there are two CEs of CE1 and CE2 connecting to the ingress RB1
associated with same pseudo-nickname, CE1 needs to locally
replicates and forwards to CE2, because another copy of the BUM
traffic between CE1 and CE2 through TRILL campus will be blocked by
the traffic filtering.
If CE1 and CE2 are not associated with same pseudo-nickname, the
copy of the BUM traffic between CE1 and CE2 through TRILL campus
won't be blocked by the traffic filtering. To avoid duplicated
traffic on receiver CE, the local replicated BUM traffic between
these two CEs on ingress RB1 should be blocked.
In summary, to ensure correct BUM traffic forwarding behavior for
each CE, local replication behavior on ingress RBridge should be
carefully designed as follows:
1. Local replication to the ports associated with the same pseudo-
nickname as that associated to the incoming port.
2. Do not replicate to mc-lag port associated with different pseudo-
nickname.
Hao & Li,etc Expires September 02, 2016 [Page 6]
Internet-Draft Centralized replication for BUM traffic March 2016
3. Do not replicate to non mc-lag ports.
The above local forwarding behavior on the ingress RB of RB1 can be
called centralized local forwarding behavior A.
If ingress RB of RB1 itself is the centralized node, BUM traffic
injected to TRILL campus won't loop back to RB1. In this case, the
local forwarding behavior is called centralized local forwarding
behavior B. The local replication behavior on RB1 is as follows:
1. Local replication to the ports associated with the same pseudo-
nickname as that associated to the incoming port.
2. Local replication to the mc-lag DF port associated with different
pseudo-nickname. Do not replicate to mc-lag non-DF port associated
with different pseudo-nickname.
3. Local replication to non mc-lag ports.
6. Loop prevention among RBridges in a edge group
If a CE sends a broadcast, unknown unicast, or multicast (BUM)
packet through DF port to a ingress RB, it will forward that packet
to all or subset of the other RBs that only have non-DF ports for
that MC-LAG. Because BUM traffic forwarding to non-DF port isn't
allowed, in this case the frame won't loop back to the CE.
If a CE sends a BUM packet through non-DF port to a ingress RB, say
RB1, then RB1 will forward that packet to other RBridges that have
DF port for that MC-LAG. In this case the frame will loop back to
the CE and traffic split-horizon filtering mechanism should be used
to avoid looping back among RBridges in a edge group.
Split-horizon mechanism relies on ingress nickname to check if a
packet's egress port belongs to a same MC-LAG with the packet's
incoming port to TRILL campus.
When the ingress RBridge receives BUM traffic from an active-active
accessing CE device, the traffic will be injected to TRILL campus
through TRILL encapsulation, and it will be replicated and forwarded
to all destination RBs which include ingress RB itself through TRILL
distribution tree. If same pseudo-nickname is used for two active-
active access CEs as ingress nickname, egress RB can use the
nickname to filter traffic forwarding to all local CE. In this case,
the traffic between these two CEs goes through local RB and another
copy of the traffic from TRILL campus is filtered. If different
ingress nickname is used for two connecting CE devices, the access
Hao & Li,etc Expires September 02, 2016 [Page 7]
Internet-Draft Centralized replication for BUM traffic March 2016
ports connecting to these two CEs should be isolated with each other.
The BUM traffic between these two CEs should go through TRILL campus,
otherwise the destination CE connected to same RB with the sender CE
will receive two copies of the traffic.
7. Centralized replication forwarding process
+-----------+
| (RB5) |
+-----------+
|
+-----------+
| (RB4) |
+-----------+
| | |
-------- | --------
| | |
+------+ +------+ +------+
|(RB1) | |(RB2) | | (RB3)|
+------+ +------+ +------+
* | * | * | ^
* | * | * | ^
* ----------*-------------*-- ^
***************************** | ^
MC-LAG1 * MC-LAG2 | ^
+------+ +------+ +------+
| CE1 | | CE2 | | CE3 |
+------+ +------+ +------+
Figure 1 TRILL Active-active access
Assuming the centralized replication solution is used in the network
of above figure 1, RB5 is the distribution tree root and centralized
replication node, CE1 and CE2 are active-active accessed to RB1,RB2
and RB3 through MC-LAG1 and MC-LAG2 respectively, CE3 is single
homed to RB3. The RBridge's own nickname of RB1 to RB5 are nick1 to
nick5 respectively. RB1,RB2 and RB3 use same pseudo-nickname for MC-
LAG1 and MC-LAG2, the pseudo-nickname is P-nick. The R-nickname on
the centralized replication node of RB5 is S-nick.
The BUM traffic forwarding process from CE1 to CE2,CE3 is as follows:
1. CE1 sends BUM traffic to RB3.
Hao & Li,etc Expires September 02, 2016 [Page 8]
Internet-Draft Centralized replication for BUM traffic March 2016
2. RB3 replicates and sends the BUM traffic to CE2 locally. RB2 also
sends the traffic to RB5 through unicast TRILL encapsulation.
Ingress nickname is set as P-nick, egress nickname is set as S-
nick.
3. RB5 decapsulates the unicast TRILL packet. Then it uses the
distribution tree whose root is RB5 to forward the packet. The
egress nickname in the trill header is the nick5. Ingress
nickname is still P-nick.
4. RB4 receives multicast TRILL traffic from RB5. Traffic incoming
port is the up port facing to distribution tree root, RPF check
will be correct based on the changed RPF port calculation
algorithm in this document. After RPF check is performed, it
forwards the traffic to all other egress RBs(RB1,RB2 and RB3).
5. RB3 receives multicast TRILL traffic from RB4. It decapsulates
the multicast TRILL packet. Because ingress nickname of P-nick is
equivalent to the nickname of local MC-LAGs connecting CE1 and
CE2, it doesn't forward the traffic to CE1 and CE2 to avoid
duplicated frame. RB3 only forwards the packet to CE3.
6. RB1 and RB2 receive multicast TRILL traffic from RB4. The
forwarding process is similar to the process on RB3, i.e, because
ingress nickname of P-nick is equivalent to the nickname of local
MC-LAGs connecting CE1 and CE2, they also don't forward the
traffic to local CE1 and CE2.
8. BUM traffic load balancing among multiple centralized nodes
To support unicast TRILL encapsulation BUM traffic load balancing,
multiple centralized replication node can be deployed and the
traffic can be load balanced on these nodes in vlan-based.
Assuming there are k centralized nodes in TRILL campus, each
centralized node has different R-nickname, VLAN-based(or FGL-based)
loadbalancing algorithm used by ingress active-active access RBridge
is as follows:
1. All R-nicknames are ordered and numbered from 0 to k-1 in
ascending order.
2. For VLAN ID m, choose the R-nickname whose number equals (m
mod k) as egress nickname for BUM traffic unicast TRILL
encapsulation.
Hao & Li,etc Expires September 02, 2016 [Page 9]
Internet-Draft Centralized replication for BUM traffic March 2016
For examples, there are 3 centralized nodes (CN) which has one R-
nickname respectively, the CN nodes will be ordered based on the
ordering algorithm for each R-nickname from CN0 to CN2. Assuming
there are 5 VLANs from 1 to 5 spreading among edge RBridges, the
traffic in VLAN 1 will go to CN1, VLAN 2 will go to CN2, and so on.
When an ingress RBridge participating active-active connection
receives BUM traffic from local CE, the RB decides to send the
traffic to which centralized node based on the VLAN-based
loadbalancing algorithm, vlan-based loadbalancing for the BUM
traffic can be achieved among multiple centralized nodes.
9. Co-existing with CMT solution
+------+ +------+
|(RB6) | |(RB7) |
+------+ +------+
------------------|-----------|----------------------
| | | | |
+------+ +------+ +------+ +------+ +------+
|(RB1) | |(RB2) | |(RB3) | |(RB4) | |(RB5) |
+------+ +------+ +------+ +------+ +------+
| | | | |
------------ -------------------------
| |
+------+ +------+
| CE1 | | CE2 |
+------+ +------+
Figure 2 CMT and centralized replication co-existing scenario
Both the centralized replication solution and CMT solution rely on
pseudo-nickname concept to avoid MAC flip-flop on remote RBridges,
these two solutions can co-exist in single TRILL network. Each
solution can be selected by each edge group RBridges independently.
As illustrated in figure 2, RB1 and RB2 use CMT for CE1's active-
active access, RB3,RB4 and RB5 use the centralized replication for
CE2's active-active access.
For the centralized replication solution, edge group RBridges should
announce local pseudo-nickname using Nickname Flags APPsub-TLV with
C-flag, the nickname with C-flag is called "C-nickname". A transit
RBridge will perform the centralized replication specific RPF check
algorithm if it receives TRILL encapsulation traffic with C-nickname
as ingress nickname.
Hao & Li,etc Expires September 02, 2016 [Page 10]
Internet-Draft Centralized replication for BUM traffic March 2016
10. Network Upgrade Analysis
Centralized nodes need software and hardware upgrade to support
centralized replication process, which stitches TRILL unicast
traffic decapsulation process and the process of normal TRILL
multicast traffic forwarding along distribution tree.
Active-active connection edge RBs need software and hardware upgrade
to support unicast TRILL encapsulation for BUM traffic, the process
is similar to normal head-end replication process.
Transit nodes need software upgrade to support RPF port calculation
algorithm change.
11. TRILL protocol extension
Two Flags of "R" and "C" in Nickname Flags APPsub-TLV [RFC7180bis]
are introduced, the nickname with "R" flag is called R-nickname, the
nickname with "C" flag is called C-nickname. R-nickname is a
specialized nickname attached on a centralized node to differentiate
unicast TRILL encapsulation BUM traffic from normal unicast TRILL
traffic. C-nickname is set on each edge group RBridge, C-nickname is
a specialized pseudo-nickname for transit RBridges to perform
different RPF check algorithm.
When active-active edge RBridges use centralized replication to
forward BUM traffic, the R-nickname is used as the egress nickname
and the C-nickname is used as ingress nickname in TRILL header for
unicast TRILL encapsulation of BUM traffic.
11.1. "R" and "C" Flag in Nickname Flags APPsub-TLV
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
| Nickname |
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
|IN|SE|R | C| RESV |
+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
NICKFLAG RECORD
o R. If R flag is one, it indicates that the advertising TRILL
switch is a centralized replication node, and the nickname is used
as egress nickname for edge group RBridges to inject traffic to
TRILL campus when the edge group RBridges use centralized
replication solution for active-active access. If flag is zero, that
nickname will not be used for that purpose.
Hao & Li,etc Expires September 02, 2016 [Page 11]
Internet-Draft Centralized replication for BUM traffic March 2016
o C. If C flag is one, it indicates that the TRILL traffic
with this nickname as ingress nickname requires special RPF check
algorithm. If flag is zero, that nickname will not be used for that
purpose.
12. Security Considerations
This draft does not introduce any extra security risks. For general
TRILL Security Considerations, see [RFC6325].
13. IANA Considerations
This document requires no IANA Actions. RFC Editor: Please remove
this section before publication.
14. References
14.1. Normative References
[1] [RFC6165] Banerjee, A. and D. Ward, "Extensions to IS-IS for
Layer-2 Systems", RFC 6165, April 2011.
[2] [RFC6325] Perlman, R., et.al. "RBridge: Base Protocol
Specification", RFC 6325, July 2011.
[3] [RFC6326bis] Eastlake, D., Banerjee, A., Dutt, D., Perlman, R.,
and A. Ghanwani, "TRILL Use of IS-IS", draft-eastlake-isis-
rfc6326bis, work in progress.
[4] [RFC7180bis] Eastlake, D., Zhang, M., Perlman, R., Banerjee, A.
Ghanwani and Gupta.S, "TRILL: Clarifications, Corrections, and
Updates", draft-ietf-trill-rfc7180bis-00, work in progress.
14.2. Informative References
[1] [RFC7781] Zhai,H., et.al., "RBridge: Pseduonode nickname",
draft-hu-trill-pseudonode-nickname-07, Work in progress,
November 2011.
[2] [TRILAA] Li,Y., et.al., " Problem Statement and Goals for
Active-Active TRILL Edge", draft-ietf-trill-active-active-
connection-prob-00, Work in progress, July 2013.
[3] [CMT] Senevirathne, T., Pathangi, J., and J. Hudson,
"Coordinated Multicast Trees (CMT)for TRILL", draft-ietf-
trill-cmt-00.txt Work in Progress, April 2012.
Hao & Li,etc Expires September 02, 2016 [Page 12]
Internet-Draft Centralized replication for BUM traffic March 2016
15. Acknowledgments
The authors wish to acknowledge the important contributions of
Hongjun Zhai, Xiaomin Wu, Liang Xia.
Authors' Addresses
Weiguo Hao
Huawei Technologies
101 Software Avenue,
Nanjing 210012
China
Email: haoweiguo@huawei.com
Yizhou Li
Huawei Technologies
101 Software Avenue,
Nanjing 210012
China
Email: liyizhou@huawei.com
Muhammad Durrani
Cisco
Email: mdurrani@cisco.com
Sujay Gupta
IP Infusion
RMZ Centennial
Mahadevapura Post
Bangalore - 560048
India
Email: sujay.gupta@ipinfusion.com
Andrew Qu
MediaTec
Email: laodulaodu@gmail.com
Tao Han
Huawei Technologies
101 Software Avenue,
Nanjing 210012
China
Email: billow.han@huawei.com
Hao & Li,etc Expires September 02, 2016 [Page 13]
| PAFTECH AB 2003-2026 | 2026-04-24 03:14:53 |