SPBm protocol as the basis of Extreme Automated Campus

    Extreme Networks Extreme Automated Campus solutions provide the ability to simultaneously use multiple transmission paths and quickly recover in the event of an accident. A network consisting of multiple physical switches is one distributed network. Traffic, at the same time, always goes along the shortest available path.
    Extreme Automated Campus is basically building an Ethernet Fabric based on the symbiosis of IEEE 802.1ah, 802.1aq and RFC6329 standards. How it works, why it is easy and convenient to install, scale and administer the open in our article.


    The “Data plane” of the factory is an IEEE 802.1ah implementation , when the ethernet packet is completely, together with the header, encapsulated in a new ethernet packet, which is why it is called MAC-in-MAC or PBB (Provider Backbone Bridges). Since MAC-in-MAC is already supported by the hardware by most of the chipsets released to the market, this, in fact, predetermined its greater popularity in comparison with the TRILL protocol.


    In the 802.1ah header, in addition to the MAC addresses and the backbone VLAN number, which are necessary for sending packets between the nodes of the factory, the I-SID (Individual Service ID) field is also transmitted, which determines the user traffic belonging to a particular service. The I-SID field has a length of 24 bit, so theoretically the factory can switch over 16.7 million different services.


    The “Control Plane” of the factory is the SPB “Shortest Path Bridging” protocol, which is approved by the IEEE 802.1aq standard . The standard itself describes two different modes of operation for SPBv and SPBm, based on VLAN and MAC-in-MAC, respectively. In our particular case, SPBm is implemented. The standard introduces a specific new terminology:


    BEB - Backbone Edge Bridge
    BCB - Backbone Core Bridge
    B-VLAN - Backbone VLAN
    C-VLAN - Customer VLAN
    UNI - User to Network Interface
    NNI - Network to Network Interface
    VSN - Virtual Service Network

    The terms themselves do not particularly need an explanation who are familiar with MPLS, they will recall terms used such as P-router, PE-router, LSP, L2 / L3VPN ...

    The SPBm logic itself is implemented using the IS-IS protocol. In RFC6329 - "the IS-the IS the Extensions Supporting the IEEE 802.1aq Shortest the Bridging the Path" describes the additional TLV necessary for correct operation of SPB.


    All switches included in the factory first establish a neighborhood using IS-IS and then form the L1 IS-IS area. Then each node calculates using the Dijkstra algorithm SPT (Shortest Path Tree) from itself to all other nodes. Then, using IS-IS, the B-VLAN FIB is filled. For B-VLAN, “flooding, broadcasting, learning” is disabled, ports cannot be added manually either. Actually, this is why the trunk MAC addresses are always known and under operator control, C-MACs are encapsulated in the 802.1ah header, and no study of these MACs takes place inside the factory.
    For Unicast, a single SPT option is calculated for each configured B-VLAN. An example of a filled Unicast FIB might look like this.

    Switch# show isis spbm unicast-fib
                              SPBM UNICAST FIB ENTRY INFO
    ADDRESS                                                INTERFACE
    00:16:ca:23:73:df   1000  0016.ca23.73df  SPBM-1        1/21            10
    00:16:ca:23:73:df   2000  0016.ca23.73df  SPBM-1        1/21            10
    00:18:b0:bb:b3:df   1000  0018.b0bb.b3df  SPBM-2        MLT-2           10
    00:14:c7:e1:33:e0   1000  0018.b0bb.b3df  SPBM-2        MLT-2           10
    00:18:b0:bb:b3:df   2000  0018.b0bb.b3df  SPBM-2        MLT-2           10
     Total number of SPBM UNICAST FIB entries 5

    For Multicast, all possible SPT pairs are calculated.


    For example, for a service terminated on 4 BEBs, 4 SPTs will be calculated for each B-VLAN. Each such SPT is calculated based on the I-SID of the configured service and BEB which is the root of this tree. Each SPBM node has its own multicast address for each VSN. This address (according to the standard) consists of 2 parts “Nickname / Source B-MAC” + “I-SID”

    For example:


    Source: 0A-BC-DE / ISID: fe-dc-ba (I-SID 16 702 650)

    (A-shifts to the “MSB” position of the top four bits; 3 - fixed value)

    Example of a filled multicast FIB:

    Switch:1(config)#show isis spbm multicast-fib
                                  SPBM MULTICAST FIB ENTRY INFO
    03:00:07:e4:e2:02  15000066 1001  0077.0077.0077 Switch-25  1/33            MLT-2
    03:00:08:e4:e2:02  15000066 1001  0088.0088.0088 Switch-33  1/50,1/33
    03:00:41:00:04:4d  1101     4058  00bb.0000.4100 Switch-1(*)1/3,1/49, TunnelHQ
    03:00:41:00:04:4f  1103     4058  00bb.0000.4100 Switch-1(*)1/3,1/49, cpp
    Total number of SPBM MULTICAST FIB entries 4

    To prevent loops, the factory uses RPFC (Reverse Path Forwarding Check), the logic of which is also provided using IS-IS. With RPFC, traffic is only received from a valid “source B-MAC / VLAN”, the remaining packets are discarded.
    Actually, after this it remains only to register the necessary services, the tincture of which is performed only on the border of the factory (that is, BEB switches).
    Examples of supported services are presented below.


    The features when working with IP multicast streams should include the following:

    • In an existing VSN, when receiving an IP multicast stream, the factory assigns a service I-SID (16000001 - 16600000) and sends this information in the form of a TLV 185/186
    • After IGMP join in this VSN, the factory calculates the SPT for the service I-SID and fills the FIB

    Advantages of Extreme Automated Campus:

    • Works on open standards
    • Supports standard Ethernet OAM tools — IEEE 802.1ag and ITU-T Y.1731
    • IS-IS works at L2 level, no IP configuration is required inside the factory
    • SPT (Shortest path tree ) is calculated on the basis of metrics and there are no blocked links
    • There is no need to configure STP in the factory core
    • All factory participants have the same SPT after calculations
    • Symmetric paths between any two nodes
    • RPFC (Reverse Path Forwarding Check) eliminates loops
    • ECMP (Equal Cost Multiple Paths) Balancing Support
    • Supported mailing multicast “1-many”, “many-1”, “many-many”
    • Factory is one virtual hop for user traffic
    • Factory operation on top of any physical topology: ring, full-mesh, partial-mesh, daisy-chain ...
    • High scalability up to 1000 nodes in one factory
    • Setting up services only at the factory border
    • Minimizing human factors
    • Convergence after failures up to

    200 ms UPD: A virtual VOSS image for testing “Extreme Automated Campus” can be downloaded on github.

    Also popular now: