Metro Ethernet Services: E-Line, E-LAN and Carrier Handoff

What is the difference between Metro Ethernet service types? We explain E-Line, E-LAN and E-Tree, VLAN and MTU settings at handoff, the effect of rate limiting on throughput and the acceptance test.
Technical summary
Technology
Metro Ethernet, MEF, 802.1Q, 802.1ad (QinQ), Y.1564
Topic
Carrier Ethernet services and handoff point design
Level
Intermediate
Prerequisites
  • Knowledge of VLANs and 802.1Q
  • Basic QoS concepts
In this article8 sections

The Metro Ethernet line has been commissioned and ping works, but file copies top out at half the contracted speed. In this picture, the cause may not be the fiber itself. The speed written in the contract is only one parameter of the service; VLAN tagging at the handoff point, the frame size that can be carried, the bandwidth profile and how layer 2 control protocols are handled determine how the line actually behaves. Below we explain carrier Ethernet services with MEF definitions and list the technical topics that should be clarified in writing before commissioning.

Metro Ethernet and MEF service types

The common language of carrier Ethernet services is set by MEF (formerly the Metro Ethernet Forum). In MEF, the physical handoff point between the customer and the carrier is called the UNI (User Network Interface), and the logical circuit connecting locations the EVC (Ethernet Virtual Connection). Service types fall into three by EVC topology; each can be offered port-based (Private) or VLAN-based (Virtual Private):

Service type Topology Port-based VLAN-based
E-Line Point-to-point EPL EVPL
E-LAN Multipoint-to-multipoint EP-LAN EVP-LAN
E-Tree Root–leaf (hub-spoke) EP-Tree EVP-Tree

In a port-based service, the entire UNI port belongs to a single EVC and customer VLAN tags are carried as is. In a VLAN-based service, there can be more than one EVC on the same physical port; the customer VLAN ID (CE-VLAN ID) determines which EVC a frame goes to. These terms don’t always appear in carrier offers in Turkey. Ask in writing which MEF type “transparent line” or “VLAN-tagged handoff” corresponds to. A verbal answer isn’t enough.

E-Line and EVPL: point-to-point

EPL behaves like a virtual cable between two locations. Customer VLANs are preserved, and both ends see the line as a single port. It’s the simplest option for a link between headquarters and a single production facility or a disaster recovery site.

EVPL, on the other hand, provides connections to multiple branches from a single physical port at headquarters. For example, on the headquarters UNI, VLAN 110 is the EVC to the first branch and VLAN 120 the EVC to the second. On the headquarters router, each VLAN becomes a separate subinterface and a separate L3 link is built with each branch. This structure also covers most E-Tree needs; branches talk to each other not directly but via headquarters.

Bandwidth profile: CIR, EIR and shaping

The MEF bandwidth profile is defined by the parameters CIR (committed information rate), CBS (committed burst size), EIR (excess information rate) and EBS (excess burst size). Traffic within the CIR is covered by the SLA; traffic falling into the EIR may be delivered but isn’t guaranteed. A situation you may encounter is a 200 Mbps service delivered over a 1 Gbps port. If the customer device pushes at 1 Gbps out of the port, the carrier’s policer drops the excess; TCP interprets these losses as congestion, shrinks the window, and throughput stays far below the contracted speed. This is the classic cause of the “copy that tops out at half speed” case from the introduction. The solution is to shape on the customer side in the egress direction according to the contracted speed. An example hierarchical policy for Cisco IOS XE:

class-map match-any VOICE
 match dscp ef
!
policy-map QOS-CHILD
 class VOICE
  priority percent 20
 class class-default
  fair-queue
!
policy-map SHAPE-200M
 class class-default
  shape average 200000000
  service-policy QOS-CHILD
!
interface GigabitEthernet0/0/1
 description METRO-ETH-UNI
 service-policy output SHAPE-200M
!
interface GigabitEthernet0/0/1.110
 description EVPL-BRANCH1
 encapsulation dot1Q 110
 ip address 10.255.10.1 255.255.255.252

The priority percentage in the example is for illustration only; calculate the size of the voice queue according to the number of concurrent calls and the codec’s bandwidth needs. Without shaping, a QoS policy is meaningless. If the queue doesn’t form on your side, priority doesn’t work on your side either.

E-LAN: multipoint L2 and its risks

E-LAN offers multipoint-to-multipoint Ethernet connectivity between more than two locations. Whether locations share the same broadcast domain depends on how you connect to this service: if you connect with a switch, the VLAN extends between locations; if you connect with a router interface, only the routers share a common transit subnet. In a design that connects with routers, the routers can form OSPF or BGP adjacencies on the common transit subnet. But weigh the risks that come with L2 between locations from the start:

  • Broadcast domain: An ARP storm, a loop or broadcast traffic produced by a faulty device at one location spreads across the whole WAN and consumes the already limited WAN bandwidth.
  • STP boundary: If location switches connect directly over the E-LAN, the STP topology extends across the WAN; a topology change at one branch can cause MAC tables to be flushed at all locations of the enterprise network infrastructure.
  • MAC limit: Carriers may limit the number of MAC addresses that can be learned per EVC; when the limit is exceeded, traffic from new devices drops and the fault comes in as “some computers can’t connect.”
  • L2 control protocols: Whether BPDU, LACP, LLDP and CDP frames are carried, dropped or processed by the carrier (MEF L2CP behavior) varies by service type and the carrier’s frame handling policy.

That’s why at each location you should connect to the E-LAN with a router or L3 switch interface; the broadcast domain and STP stay within the location. A single loop takes down all branches. For STP design options, see the STP, RSTP and MSTP comparison article.

The handoff point: VLAN tagging, MTU and QinQ

Don’t accept the line until three topics are clear at the handoff point.

Tagging: The carrier can hand off the UNI untagged, with a single 802.1Q tag, or carrying customer tags as is. In a port-based service, the customer can send whatever VLANs it wants; in a VLAN-based service, only the agreed VLAN IDs are accepted and the others are silently dropped. Forgetting to notify the carrier when you add a new VLAN is a cause of “the line isn’t working” you may encounter.

QinQ: To separate customer traffic in its own network, the carrier adds a service tag (S-tag) on top of the customer tag. IEEE 802.1ad defines the TPID value 0x88A8 for this outer tag; some carriers still use 0x8100. If a double-tagged handoff will be made to the customer device, the TPID value and tag order should be confirmed in writing.

MTU: An untagged Ethernet frame carrying a 1500-byte IP packet is 1518 bytes including the FCS; with a single 802.1Q tag it becomes 1522, and with double tags 1526 bytes. If IPsec, GRE or MPLS will be carried over the line, additional headers are added. The maximum frame size the carrier supports (for example, jumbo support at the 1600 or 9000-byte level) should be written in the contract. Otherwise large packets are silently dropped and the problem shows up only in certain applications: ping works, login works, file sharing or replication stalls. That’s what an MTU problem looks like.

L2 or L3 between locations?

The default choice is L3: each location has its own subnets, traffic between locations is routed, and the broadcast domain and failure impact stay limited to the location. How to structure the enterprise VLAN plan per location is explained in the VLAN design article.

Consider L2 extension only if there’s a real requirement: an old application that has to stay in the same subnet, virtual machines that need to change location without changing IP, or industrial systems that communicate only at L2. In that case, minimize the number of extended VLANs, enable storm control and BPDU protection on edge ports, and get the MAC limit and L2CP behavior from the carrier in writing. L2 isn’t extended “just in case.”

Service verification and the acceptance test

Accepting a circuit because “it pings” doesn’t show that the contracted parameters are met. Many carriers do a service activation test based on ITU-T Y.1564 at commissioning; this test measures throughput, latency, latency variation and frame loss together at the CIR and EIR levels. Request the test report and do these checks on your side:

  1. A bidirectional throughput test (for example, with iperf3 at both ends, with multiple parallel streams). iperf3 measures throughput achieved at the application level; the Ethernet service acceptance criteria in the contract (committed rate, frame loss, latency) are evaluated with the carrier’s service acceptance test. Record the two results separately. A single stream can give a misleading result because of the window size.
  2. A maximum unfragmented packet size test.
  3. Verifying that all agreed VLANs are carried end to end and that other VLANs are dropped.
  4. Recording a baseline measurement for latency and jitter; it becomes the reference in later fault reports.

Without a baseline measurement, a fault ticket turns into an argument. Continuous measurement of line quality after commissioning and proving SLA violations are the subject of the WAN SLA measurement article. If the carrier does OAM monitoring with Y.1731 or 802.1ag CFM, also ask whether you can access this data as a customer.

Checklist of technical questions to ask the carrier

  • What is the service type in MEF terms? (EPL, EVPL, EP-LAN, EVP-LAN)
  • UNI physical interface: copper or fiber, speed, SFP type and wavelength?
  • Is the handoff untagged, 802.1Q or double-tagged? What is the TPID of the outer tag?
  • Which VLAN IDs will be carried, and is the CE-VLAN ID preserved?
  • What is the maximum frame size in bytes?
  • What are the CIR, CBS and EIR values, and what behavior is applied to excess traffic (drop or mark)?
  • Are there QoS classes and a DSCP/PCP mapping in the carrier backbone?
  • Is there a MAC limit per EVC?
  • Are STP BPDU, LACP, LLDP and CDP frames carried or dropped?
  • How are latency, jitter, loss and availability SLA values measured, and how is the report shared?
  • The fault reporting channel and response time; how far in advance is planned maintenance announced?
  • Physical route: does the backup line come from a different exchange and building entrance?

The last item is the one most often skipped. Two lines coming through the same duct go down together with a single excavator strike. If you haven’t chosen the WAN model yet, first read the SD-WAN and MPLS comparison; where Metro Ethernet fits in the architecture becomes clear. For a technical assessment of existing lines and carrier coordination, see the Metro Ethernet and internet line management page.

Frequently asked questions

What is the difference between Metro Ethernet and MPLS?

Metro Ethernet provides the customer a layer 2 Ethernet service; routing is on the customer side; whether locations share the same broadcast domain depends on the service type and how the customer connects. MPLS L3VPN, on the other hand, is a layer 3 service; routing is shared with the carrier. Carriers often carry Metro Ethernet over MPLS in their own backbone, but the service type visible to the customer is different.

Should you choose EPL or EVPL?

If you’ll connect to a single remote location and use the whole port, EPL is simpler and carries customer VLANs transparently. If separate circuits to multiple locations will be connected through a single physical port at headquarters, EVPL is suitable; each circuit is separated with its own VLAN.

How does the carrier limit bandwidth?

Usually with a policer in the ingress direction that marks or drops frames based on the CIR and, if applicable, the EIR. If the port speed is higher than the contracted speed, excess traffic is dropped on the carrier side and TCP throughput collapses. To prevent this, do shaping in the egress direction on the customer device according to the contracted speed.

Should I use STP on an E-LAN service?

If possible, don’t extend STP between locations. Connecting to the E-LAN with a router or an L3 interface at each location keeps the broadcast domain and STP boundary within the location. If L2 extension is mandatory, clarify the carrier’s BPDU handling behavior and enable protection mechanisms on edge ports.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Network

We build switching, routing, cabling and wireless networks with VLAN segmentation and redundant uplinks, and provide regular maintenance and fault support.

Knowledge Center

Other articles