Testing Juniper Networks M40 Router MPLS InterOperability with Cisco Systems 7513 and 12008 Routers Mark Anderson, Systems Engineer Annette Kay Donnell, Marketing Engineer Juniper Networks, Inc.
Executive Summary
Juniper Networks, Inc. condUCted a series of interoperability tests at our facilities. The purpose is to demonstrate that the MPLS implementation of the M40 Internet router interoperates with Cisco Systems 7513 and 12008 routers. You can use the PRocedures in this paper to replicate the tests. The interoperability test includes the creation and use of label switched paths (LSPs), including the control plane using the Resource Reservation Protocol (RSVP) to setup LSPs and the forwarding plane that forwards packets through LSPs. It also includes the testing of IS-IS traffic engineering extensions. The tests were successful in that they demonstrated that Juniper Networks and Cisco Systems MPLS implementations are interoperable.
Scope
Appendix A lists the specific configuration used for all three configuration tests. This configuration enables you to test the MPLS interoperability between the M40 router running JUNOS Internet software release 3.4, Cisco Systems 7513 routers running IOS 12.0(6)S, and 12008 routers running IOS 12.0(7)S. Particularly, these tests verify LSPs and RSVP, as well as IS-IS traffic engineering extensions. Features not tested include Juniper Networks enhancements, such as MPLS Fast Reroute, Circuit Cross Connect, and OSPF traffic engineering. Other features not tested include the set of tag-switching features that Cisco Systems calls MPLS features, such as MPLS VPN.
These tests do not include performance testing. Test Environment
The test environment consists of a four-router network topology: two M40 routers, one Cisco Systems GSR 12008, and one Cisco Systems 7513 (Figure 1).
Figure 1: MPLS Interoperability Test Topology
点击查看大图
LSP Interoperability Test
This test determines whether you can create an LSP across an arbitrarily heterogeneous path, as well as whether static routing correctly operates.
Parameters Appendix A lists the configurations for the four routers in the LSP and static routing tests.
To demonstrate interoperability over an arbitrary path, a set of permutations of possible paths is created (Table 1). This set represents all the possible paths. Each member of the set contains a different combination of routers for three different label switching router (LSR) functions: ingress, transit, and egress. These six paths are overlaid on top of the topology in Figure 2.
Table 1: LSP Created Ingress Transit Egress Cisco Systems Cisco Systems Juniper Networks Cisco Systems Juniper Networks Juniper Networks Juniper Networks Cisco Systems Juniper Networks Cisco Systems Juniper Networks Cisco Systems Juniper Networks Juniper Networks Cisco Systems Juniper Networks Cisco Systems Cisco Systems
点击查看大图
To test different means of path selection, the LSPs in Figure 2 are configured in a variety of ways. Most LSPs are configured with eXPlicit paths to constrain them to a single path. Others, like the C2-J2-J1 LSP, are configured to dynamically choose a path that meets the bandwidth requirements of the LSP.
IP traffic is sent over all the LSPs using a Smartbits (SMB) 6000. The SMB 6000 is connected through a Gigabit Ethernet port directly to each router. To route IP traffic through the LSPs, each Gigabit Ethernet circuit is advertised via BGP.
Results Testing demonstrates that the MPLS LSP implementations and static routing implementations are interoperable.
For each Gigabit Ethernet circuit, the BGP next hop advertises for the loopback interface of the router connected to it. Likewise, every LSP is established between the loopback interfaces of the originating and terminating routers. Any LSP originating on an M40 router dynamically uses this information to route traffic over the LSP. When a Cisco Systems-initiated LSP is configured with autoroute announce, the same behavior occurs on the Cisco Systems router.
Implementation Differences The two implementations differ in how traffic is routed through the LSP.
In JUNOS software, there are three ways to route traffic into LSPs. The primary means is with BGP. If any BGP route exists using the LSP destination as the BGP next hop, then BGP automatically uses the LSP to reach the next hop instead of using IGP.
JUNOS software enables you to use IGP next hops in addition to BGP next hops to route traffic through LSPs.
You can statically map routes into an LSP.
In Cisco Systems IOS, there are two ways to route traffic into LSPs: static routes and automatic routes. Cisco Systems' Autoroute feature sends all traffic to a specific destination through the same LSP and routes to all the IGP interfaces on the destination router. There is no Cisco Systems equivalent to the JUNOS default of using solely BGP to place routes through the LSP.
RSVP Interoperability
By successfully creating and configuring the LSPs in the above LSP test, the interoperability of RSVP signalling is tested indirectly.
Parameters Appendix A lists the configurations for the four routers in the RSVP test.
Results Testing demonstrates that the RSVP implementations are interoperable. Specifically, the following RSVP objects are interoperable between Juniper Networks and Cisco Systems routers.
Explicit-Route Label Label-Request session The only object not interoperable is the Record-Route object. This exception, which is being addressed by both Juniper Networks and Cisco Systems, does not affect the creation or use of LSPs.
Implementation Differences Juniper Networks supports the RSVP Hello packet and Cisco Systems does not. This difference does not prevent Cisco Systems from creating and using LSPs.
To signal penultimate hop popping, a Cisco egress router sends label 0 to the penultimate hop in the control plane, whereas a Juniper Networks egress router sends label 3 (implicit NULL). When a Cisco egress router is connected to a penultimate Juniper Networks router, it is recommended that you use the following IOS command to require the egress router to send label 3 (as defined by the standard).
The inspection of the Traffic Engineering Database (TED) during the LSP test indicates whether Cisco Systems and Juniper Networks are sending the correct information in the sub-TLVs in the new IS-IS extended IS reachability TLV type 22. The identical use of the information in the TED is established by adding a few more steps to the existing test as described in the Parameters section.
Parameters Appendix A lists the basic configurations for the four routers in the tests, though the following changes are made.
Bandwidth Reservation Some additional LSPs are created between J1 and C2 to determine whether the routers in the path correctly update the unreserved bandwidth in the sub-TLV. The configuration of one of these LSPs is as follows.
