Metro Ethernet LAN Services Overview
Multiple services can be implemented and offered by ISPs for the MAN connectivity. Services can be Layer 1 to Layer 3 based and operates as P2P or P2M. The following pictures resumes the deployment options :
Ethernet Private Line (EPL)
Port-based circuit P2P Ethernet mapping L2 traffic on a TDM Circuit. This is a P2P connection from one location to another with guaranteed bandwidth and payload transparency.
This typically use SONET/SDH as transport. The bandwidth is guaranteed and dedicated with no oversubscription. An SLA concerning uptime may support all requirements. SONET protection can add HA to this service. EPL is used for transparent LAN and Datacenter integration where wire speed and VLAN transparency is needed.
EPL also supports DWDM/CWDM, EoSONET. This is a layer 1 service where multiple customers can have dedicated SONET channels over a mutualized SONET infrastructure.
Ethernet Relay Service (ERS)
Point to Point VLAN-Based service supporting multiplexing where multiple EVC (Ethernet Virtual Connection) can be multiplexed over a single customer UNI (User Network Interface).
Multiple connections can be provided one only one link and the multiplexed UNI can support P2P or P2M between multiple sites (like FR). The connection ID, instead of being a DLCI is a VLAN tag. Each customer tag is mapped to an EVC. The VLAN tag indicates the destination, the Ethernet service is not L2 transparent. ERS will use different P2P VLANs to connect one site to other remote sites.
ERS provides Ethernet access with service networking to other L2 services like Frame Relay or ATM. Traffic on a DLCI or VP/VC is converted to an Ethernet frame by the SP. This allow customers to use Ethernet Services without replacing their legacy connections.
Providers can offer tiered services based on Bandwidth, CoS or Distance. SLA may be based on CIR or PIR, burst and packet-loss rate. Providers can multiplex connections from many end customers onto a single Ethernet port on a POP. When ERS is implemented over MPLS, there is 1:1 mapping between VLAN IDs and EoMPLS pseudowires.
Ethernet Wire Service (EWS)
EWS is a P2P between a pair of site. The difference with EPL is that EWS is usually provided over a shared and switched infrastructure within the SP network. Oversubscription is handled using stat mux.
EWS customers have the choice between multiple level of service from committed bandwidth up to wire speed. To keep privacy, SP will segregate traffic by applying VLAN tags on each EVC (with QinQ for example). SLA are usually based on CoS. EWS is considered port-based service providing one to all bundling. The CE can be a router or a switch.
EWS is commonly used for P2P Lan Extension, access to storage or Datacenter connectivity. EWS provides Layer 2 transparency.
Ethernet Multipoint Service (EMS)
EMS is M2M provided over a shared, switched infrastructure. This a the multipoint version of EWS and it also can be enabled by using VPLS at the NPE. The provider will segregates each customer’s traffic by applying VLAN tag on each EVC typically using QinQ. EWS offers the same feature as EWS but in a multipoint fashion.
Ethernet Relay Multipoint Service (ERMS)
ERMS is a hybrid of EMS and ERS. It offers any to any connectivity like EMS and service multiplexing of ERS. It’s an Ethernet M2M VLAN-based service targeted with Layer 3 devices at CEs. It is useful when customers needs M2M connection among WAN routers.
This allows on UNI to support a customer’s intranet connection (site to site) and one or more additional EVCs for connection to outside network like ISPs or content provider.
ATOM (Any Transport Over MPLS)
MPLS is the center technology inside ISPs Core networks. MPLS networks were deployed to supported IP services such as Layer 3 MPLS VPNs. When Layer 2 services were required, separated Layer 2 networks were built over Layer 1 infrastructure like SONET/SDH, CWDM/DWDM. ATOM allows ISP to use the existing MPLS network as a transport network for Metro Ethernet (Layer 2 Based) services.
ATOM currently support transport of Ethernet, HDLC, PPP, Frame Relay and ATM. ATOM can also transport TDM based services over an MPLS network. Mixing different types of transport is also supported.
To build ATOM, Layer 2 P2P pseudowires are used between two CE. This pseudowire is presented as a regular layer 2 circuit to the CE and frames are encapsulated with ATOM header specific to the Layer 3 Protocol.
Most of the time, ATOM is used to interconnect to identical type of connection like EoMPLS connecting to Ethernet Segment or ATM over MPLS connecting two ATM segment.
Ethernet Over MPLS
EoMPLS is a layer 2 transport option that ATOM is able to offer.
An EoMPLS Virtual Circuit is signaled and established through a targeted LDP session. This is a point to point connection so MAC learning is not necessary on these links. The EoMPLS circuit represents a virtual PW. When an Ethernet frame is received from the CE router/switch, the PE strips the preamble and FSC and add the following headers :
- ATOM Control Word : Optionnal for Layer 2, except for FR and ATM AAL5. This is a 32 bit header used to transport information such as protocol control and sequence number. Its usage (or not) is negotiated when the VC is established.
- Circuit Label : Inner MPLS Label representing the circuit. This label is not used in the Core and is only used by egress/ingress PE to determine on which interface the frame needs to be forwarded.
- Tunnel Label : This is the MPLS Label used in the Core to switch the label to the remote PE. This label is used through the normal LDP behavior or could be determined with RSVP-TE.
- Core Layer 2 Header : This header is used by the egress PE to encapsulate the EoMPLS frame in the appropriate Layer 2 protocol (Frame Relay, ATM or Ethernet…) and sends it to the next-hop router in the LSP.
EoMPLS can be used in Port Mode or in VLAN Mode. The mode determines how 802.1Q tags are processed :
Port mode is also called Circuit type 5. In this mode the entire Layer 2 frame received in the ingress interface is encapsulated including 802.1Q header. This allows transparent service to the customer and allows to use the pseudowire as a standard Ethernet Link. Multiple VLANs can be carried, EoMPLS Port Mode implement an EWS.
Circuit Mode or Circuit type 4 use the VLAN tag to determine the circuit where the frame needs to be transmitted. The VLAN Tag acts as a Circuit ID like a DLCI or VP/VC. At the egress port, the 8021.Q tag is rewritten to the VLAN that identifies the circuit on the other end and this can be a different VLAN. Customer VLAN are not transmitted transparently and the 802.1Q header is mandatory to select a destination. EoMPLS implements an ERS.
How to choose
To choose between all these connectivity options, global needs must be found. Is there PDU transparency needed for L2 protocols like STP, is there anything like VMotion needed ? Is multipoint needed or a collection of point to point links are ok ?
VPLS is a multipoint architecture connecting two or more customers devices using Ethernet bridging over MPLS. P network emulates a 802.1D bridge with each EMS acting like a VLAN. There is two RFCs incompatible with each other, the first one is proposed by Juniper and use BGP Auto Discovery and Signaling and the second one proposed by Cisco that use LDP Signaling.
UPE devices act as 802.1D bridges. They are connected in a full mesh fashion with PWs that are used over an MPLS Core. From the UPE perspective, PWs are just regular Ethernet connections.
VPLS will use the MAC learning process to learn the MAC and their port association. As in regular Ethernet networks, if a destination MAC is not known, broadcast or multicast, they frame will be flooded out all ports associated with the virtual bridge. The VPLS Core does not use STP and rely on split horizon so that frames are not forwarded out the PWs they were received.
Standard implementation of VPLS requires a full mesh of PWs. The issues begins when the number of device increase. H-VPLS offers a scaling option that requires a full mesh of PWs only for the VPLS core.
When PWs redundancy is implemented, the UPE will block all PWs except the one that goes to the highest IP address. H-VPLS offers the possibility to have the MPLS Core to also serve for L3 VPN. MPLS Core will constrain STP to smaller domain, speeding up convergence and limiting errors spreading. H-VPLS provides flexible model that enables Multipoint Ethernet services, Point to point Layer 2 VPN service and Ethernet access to Layer 3 VPN services.