2.802.1q SPB
Last updated
Was this helpful?
Last updated
Was this helpful?
SPB can co-exists with existing 802.1d and 802.1q bridged lan, still can be an augment to bridged LAN. SPB employs PBB(provider backbone bridge) as the encapsulation ideology ,it offers customer bridging service ,provider bridging service and provider backbone bridging service to customers, while TRILL is customer bridging only.
same with TRILL ,the core bridging does not care end station at all,thus that the core can scale with the number of core bridges.
though it is claimed routing based switching ,it does not contain any headers which can convey the basis for routing and forwarding .it's is very implicit.as for routing and loop mitigation ,it's not as good as TRILL.
In SPB ,the path ,unicast or multicast ,is symmetric ,and it can support ECMP for different traffic from different tenants,it does not support sharing traffic from same tenant among ECMP .
next we should SPB has advantage over TRILL in multi-tenancy ,because of PBB is natively supported.and it could EVPN can be self-serviced,which means we can allocate E-LINE ,E-LAN, E-TREE service from PBB natively.
if you'v once known VXLAN which is an overlay method which encapsulates L2 frame over L4 UDP ,you can easily what's PBB.
it's also an overlay technology ,often we say it's a MAC-in-MAC encapsulation,you know vxlan relies on L3 protocol stack routing and forwarding,when it comes to mac-in-mac ,L2 process just switches them according to outer Ethernet addresses. with SPB ,routing is also involved.
PBB provides another layer to segregate customer away from backbone provider,that's to say ,the core does not need to know what customer frame exactly is,we first represent the frame structure below:
full image:https://chillancezen.gitbooks.io/protocol-and-specification/content/assets/import12121.png
note there may be 4 tags in a PBB frame ,they are B-tag,I-tag,S-tag,C-tag, B-tag is the backbone vlan tag which can divide underlay network into more than one logical network,usually there will be a mapping between I-tag and B-tag ,1-1 or n-1 .I-tag is the instance tag, similar to VXLAN header in vxlan frame,it contains a 24-bit(16M) I-SID space,it could support almost 16M tenants(which has the same tenancy capability as with VXLAN ,but it runs directly on top of Layer 2).S-tag is 802.1ad related vlan tag(also the QinQ outer tag),PBB is originally designed to mitigate limited vlan range in QinQ ,and it could recognize QinQ traffic and provide them backbone bridging service .last C-Tag is for common customer traffic .no matter if traffic is vlan tagged .
as a matter of fact ,PBB only cares how to encapsulate/decapsulate a packet,and define backbone bridge domains,and also bridge domain.
the following paragraphs will tell what components a PBB bridge should have ,but does not reveal a specific implementation.
In TRILL ,a network over which RBridges are deployed is called a campus . similarly ,we call this network PBBN.
a bridge which is positioned at core and act the role of transit rbridge in TRILL is called BCB(Backbone Core Bridge), the bridged positioned at the edge is called BEB(Backbone Edge Bridge),a BEB receives customer traffic and encapsulate them by PBB frame format.
there are two kinds of component in BEB or BCB:
I-Component :an entity that encapsulation traffic from CNP(Customer Network Port) with a I-tag and transmit them through PIP(Provider Instance Port),there is a mapping for S-tag(or port-id or C-Tag) and I-tag which,once the packets finishes I-Tag encapsulation,it choose a PIP port ,send it out.note that PBB specification does not specify any method to transmit a I-taged packets ,I guess maybe a tunnel can do it.PBB say its PIP should terminate SPT from core PBBN.
B-component :an entity which is responsible for encapsulating received packets from CBP(Customer Backbone Port) and then add B-tag as well as outer Ethernet ,throw them through PNP(Provider Backbone Port) into PBBN core network.in addition ,CBP should terminate any BPDU traffic from customer edge.
A BEB bridge can contain I-component or B-Component or both components,it providers several customer service interfaces -- \_Port Based,S-tagged and I-tagged_ ,the following figure depict what comprises a BEB:
the CNP decides in what way a packet can access PBBN,for example ,a port based CNP will treat all the traffic from a specific port as traffic from an instance .
SPB can run on two modes :SPBV and SPBM.
in SPBV s-vlan or c-vlan within customer frame is translated into SPBVID which corresponds to a specific SPT in PBBN,PBBN does not modified the frame ,instead ,addresses from all stations must be learned.
when SBP is combined with PBB, SPBM comes ,PBB contains 24-bit I-SID,so 2^24 different backbone services can be configured,each I-SID is mapped to a backbone VID(B-VID),'backbone vlan must be present.
what comprise PBB outer mac addresses? you may wonder this. ingress BEB's PNP's mac will be filled into PBB's outer source address,and egress BEB's PNP's mac is filled into the outer destination address.thus the packet then can be routed correctly.
In PBB ,BCB and BEB's B-component must be part of the whole big layer 2 network, that's to say,every BEB or BCB should learn mac addresses of existing bridges within a specific VLAN(configured on a PNP) ,and implement at least part of L2 process features .
one more thing is existing infrastructure network should be vlan aware (it recognizes VLAN and forwards packets on a per-vlan basis).
we know now ,SPB is kind of overlay networking when PBB is used ,it provides multi-tenancy and scales well.but it's not infrastructure friendly as TRILL(Where network is organized into campus,and it explicitly use L3 routing) is.
may be we can combine TRILL and SPBM to overcome the disadvantages each of them individually brings.
