Network Working Group | Z. Chen |
Internet-Draft | Huawei |
Intended status: Standards Track | X. Xu |
Expires: September 6, 2018 | Alibaba |
D. Cheng | |
Huawei | |
March 5, 2018 |
Overheads Reduction for IS-IS Enabled Spine-Leaf Networks
draft-chen-isis-sl-overheads-reduction-03
When a Spine-Leaf topology adopts the Intermediate System to Intermediate System (IS-IS) routing protocol, the Leaf node receives Link State Packets (LSPs) from all the other nodes thus having the entire routing information of the topology. This is usually considered unnecessary and costly. This document describes a solution to this problem by utilizing IS-IS's inherent multi-level and area partition features, which requires that an IS-IS router SHOULD check a level-1 LSP's area addresses before advertising it to a neighbor.
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.
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 September 6, 2018.
Copyright (c) 2018 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 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.
Spine-Leaf topology (a.k.a., CLOS topology) is widely used in today's datacenter and campus networks. When the Spine-Leaf topology runs the Intermediate System to Intermediate System (IS-IS) routing protocol, each Leaf node receives Link State Packets (LSPs) from all the other nodes thus having the entire routing information of the topology. This is usually considered unnecessary and costly because the Leaf node only needs to know its default gateways (i.e., the Spine nodes it connects to) and the LSPs generated by the other Leaf nodes bring little benefit for it to forward traffic.
To avoid Leaf nodes from learning the unnecessary LSPs from one another, [IS-IS-SL-Extension] proposes a new TLV attached to the IS-IS Hello (IIH) PDU to carry an router's role (i.e., Spine or Leaf) in the topology. The Spine nodes then prevent all LSPs from being sent to the Leaf nodes, and each Leaf node sets the Spine nodes it connects to as its default gateways.
This document proposes another solution to this problem, which utilizes IS-IS's inherent multi-level and area partition features. In particular, it requires that each Leaf node (configured as L1 router) SHOULD be assigned with a unique area address and each Spine node (configured as L1/L2 router) MUST NOT advertise level-1 LSPs of a given area to neighbors within another area. This prevents Leaf nodes from receiving routing information from one another, without introducing new message formats.
+------------+ +------------+ | Spine-A | 10.10.10.0/24 | Spine-B | | L1/L2 +----------------------+ L1/L2 | | Area10/20 | .1 .2 | Area10/20 | +---+--+-----+ +---+----+---+ .1 | | .1 .2 | | .1 | | 10.10.40.0/24 | | | | +-----------------------------+ | 10.10.20.0/24 | | | | 10.10.30.0/24 | +--|-------------------------------+ | | | 10.10.50.0/24 | | .2 | | .1 .2 | | .2 +---+-----+--+ +-----+--+---+ | Leaf-A | | Leaf-B | | L1 | | L1 | | Area10 | | Area20 | +-----+------+ +-----+------+ | | | | ------+------- ------+------- 192.168.10.0/24 192.168.20.0/24 Figure 1: Topology Example
To prevent a Leaf node from learning the routing information of the other ones, the following configurations are REQUIRED:
As a result, Leaf-A and Leaf-B in Figure 1 are configured as L1 routers and are assigned 10 and 20 as their area addresses, respectively. Spine-A and Spine-B are configured as L1/L2 routers and are assigned both 10 and 20 as their area addresses.
Level-1 Link State Database (Spine-A): +--------------------+----------+--------+--------+------+--------+-----+ |LSPID |Seq Num |Checksum|Holdtime|Length|ATT/P/OL|Area | +--------------------+----------+--------+--------+------+--------+-----+ |1111.1111.1111.00-00|0x0000006c|0x540b |743 |124 |0/0/0 |10/20| +--------------------+----------+--------+--------+------+--------+-----+ |2222.2222.2222.00-00|0x0000006d|0x933b |1068 |124 |0/0/0 |10/20| +--------------------+----------+--------+--------+------+--------+-----+ |3333.3333.3333.00-00|0x0000006b|0x1815 |402 |122 |0/0/0 |10 | +--------------------+----------+--------+--------+------+--------+-----+ |4444.4444.4444.00-00|0x0000006a|0xf543 |431 |122 |0/0/0 |20 | +--------------------+----------+--------+--------+------+--------+-----+ Level-2 Link State Database (Spine-A): +--------------------+----------+--------+--------+------+--------+-----+ |LSPID |Seq Num |Checksum|Holdtime|Length|ATT/P/OL|Area | +--------------------+----------+--------+--------+------+--------+-----+ |1111.1111.1111.00-00|0x0000006f|0x682f |743 |150 |0/0/0 |10/20| +--------------------+----------+--------+--------+------+--------+-----+ |2222.2222.2222.00-00|0x00000063|0x30eb |1068 |150 |0/0/0 |10/20| +--------------------+----------+--------+--------+------+--------+-----+ Figure 2: Link State Database of Spine-A
Under such configurations, however, Leaf-A still receives Leaf-B's LSPs (and vice versa) even though they are in different areas. This is because of the IS-IS definition that all routers in a specific area SHOULD share the same level-1 Link State Database (LSDB). In other words, IS-IS routers check area addresses during neighbor establishment, but are regardless of area addresses when advertising LSPs to a neighbor.
The example in Figure 1 and the LSDB of Spine-A (in Figure 2) further illustrate this. Since Spine-A and Leaf-B are both in area 20, Spine-A will receive LSP 4444.4444.4444.00-00 from Leaf-B and store the LSP into its level-1 LSDB. On the other hand, since Spine-A and Leaf-A are both in area 10, Spine-A will advertise LSP 4444.4444.4444.00-00 to Leaf-A although Leaf-A and Leaf-B (generator of the LSP) are in different areas. As a result, Leaf-A installs the route 192.168.20.0/24 into its routing table (Figure 3), even though it is an external area route.
Leaf-A Routing Table: +---------------+-------+---+----+-----+----------+--------------+ |Destination |Proto |Pre|Cost|Flags|NextHop |Interface | +---------------+-------+---+----+-----+----------+--------------+ |10.10.10.0/24 |ISIS-L1|15 |20 |D |10.10.20.1|Ethernet0/0/0 | | |ISIS-L1|15 |20 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.20.0/24 |Direct |0 |0 |D |127.0.0.1 |Ethernet0/0/0 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.30.0/24 |ISIS-L1|15 |20 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.40.0/24 |Direct |0 |0 |D |127.0.0.1 |Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.50.0/24 |ISIS-L1|15 |20 |D |10.10.20.1|Ethernet0/0/0 | +---------------+-------+---+----+-----+----------+--------------+ |192.168.10.0/24|Direct |0 |0 |D |127.0.0.1 |GEthernet0/0/0| +---------------+-------+---+----+-----+----------+--------------+ |192.168.20.0/24|ISIS-L1|15 |30 |D |10.10.20.1|Ethernet0/0/0 | | |ISIS-L1|15 |30 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |127.0.0.0/8 |Direct |0 |0 |D |127.0.0.1 |InLoopBack0 | +---------------+-------+---+----+-----+----------+--------------+ |0.0.0.0/0 |ISIS-L1|15 |10 |D |10.10.20.1|Ethernet0/0/0 | | |ISIS-L1|15 |10 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ Figure 3: Routing Table of Leaf-A
Therefore, the solution proposed in this document requires that an IS-IS router SHOULD check a level-1 LSP's area addresses before advertising it to a neighbor (see Section 2.2).
Before advertising a level-1 LSP to a neighbor, an IS-IS router SHOULD compare the area addresses associated with the LSP and the ones associated with the neighbor. If they have at least one area address in common, the router SHOULD advertise the LSP to the neighbor. Otherwise, the router MUST NOT advertise the LSP to the neighbor.
In the former case, the router SHOULD remove every area addresse in the LSP except the ones associated with the neighbor before the advertisement. This makes the solution more compatible since the Leaf nodes can be unaltered (see Section 3.2).
For instance, before Spine-A advertises LSP 1111.1111.1111.00-00 to Leaf-A, it compares the LSP's area addresses (i.e., 10 and 20) with Leaf-A's area address (i.e., 10). Since they have a common area address 10, Spine-A SHOULD remove area address 20 from the LSP and advertise the LSP to Leaf-A. On the other hand, before Spine-A advertises LSP 4444.4444.4444.00-00 to Leaf-A, it checks their area addresses and finds that they have no area address in common. So Spine-A MUST NOT advertise the LSP to Leaf-A. As a result, Leaf-A would not learn any routing information of Leaf-B, as shown in Figure 4.
Leaf-A Routing Table: +---------------+-------+---+----+-----+----------+--------------+ |Destination |Proto |Pre|Cost|Flags|NextHop |Interface | +---------------+-------+---+----+-----+----------+--------------+ |10.10.10.0/24 |ISIS-L1|15 |20 |D |10.10.20.1|Ethernet0/0/0 | | |ISIS-L1|15 |20 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.20.0/24 |Direct |0 |0 |D |127.0.0.1 |Ethernet0/0/0 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.30.0/24 |ISIS-L1|15 |20 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.40.0/24 |Direct |0 |0 |D |127.0.0.1 |Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ |10.10.50.0/24 |ISIS-L1|15 |20 |D |10.10.20.1|Ethernet0/0/0 | +---------------+-------+---+----+-----+----------+--------------+ |192.168.10.0/24|Direct |0 |0 |D |127.0.0.1 |GEthernet0/0/0| +---------------+-------+---+----+-----+----------+--------------+ |127.0.0.0/8 |Direct |0 |0 |D |127.0.0.1 |InLoopBack0 | +---------------+-------+---+----+-----+----------+--------------+ |0.0.0.0/0 |ISIS-L1|15 |10 |D |10.10.20.1|Ethernet0/0/0 | | |ISIS-L1|15 |10 |D |10.10.40.2|Ethernet0/0/1 | +---------------+-------+---+----+-----+----------+--------------+ Figure 4: Routing Table of Leaf-A
As defined in [RFC 1195], a L1/L2 router will indicate in its LSPs that it is "attached" by setting the ATT bits. Therefore, each Leaf node would set the Spine nodes as its default gateways and install a default route in its routing table, as shown in Figure 4.
However, a specific IS-IS implementation in this case may not let the L1/L2 router set the ATT bits, because it may speculate that the L1/L2 router has lost connectivity to the level-2 backbone. To solve this problem, operators can manually configure the L1/L2 router to advertise a default route.
In most deployments, an IS-IS router is assigned only one area address, which will not be influenced by the area checking mechanism proposed in this document. However, an IS-IS router might be assigned more than one area addresses in some practical deployments for the following reasons: 1) it is desirable to change the area address of an area, 2) to merge two areas into one area, or 3) to partition an area into two areas.
For instance, to change an area's address from X to Y, one can simply add area address Y to all routers in the area, and then remove X from them. Note that such operations would not disrupt live traffic in the network.
Although the solution in this document requires IS-IS router to check LSP's area addresses before advertising it, the above use cases are still applicable and no compatible issue rises.
The maximumAreaAddresses parameter in today's IS-IS implementation is set to be 3 (or 0 which indicates 3) on consensus. Therefore, the solution in this document also requires that Spine node SHOULD be modified for supporting more area addresses. However, as LSPs sent to a given neighbor only carry the area address(es) of the neighbor (see Section 2.2), the solution does not require to modify Leaf nodes.
TBD.
TBD.
TBD.
[IS-IS-SL-Extension] | Shen, N., Ginsberg, L. and S. Thyamagundalu, "IS-IS Routing for Spine-Leaf Topology", draft-shen-isis-spine-leaf-ext-05 (work in progress) , January 2018. |
[RFC1195] | Callon, R., "Use of OSI IS-IS for Routing in TCP/IP and Dual Environments", RFC 1195 , December 1990. |
[RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |