Carrier Ethernet
From its modest beginning in the early 1970s as a way to connect computers in a
data center, Ethernet has evolved over the years into the dominant networking
technology. Today it is used in a wide range of environments, from home networks
connecting a few computers to large enterprise networks connecting tens of
thousands of nodes. While it was originally designed as a local area networking
technology, it is now increasingly used by telecommunications carriers as an
alternative to traditional circuit switched networks over long distances. Carrier
Ethernet refers to the extensions necessary in order to allow telecommunications
carriers to deploy and provide Ethernet services to their customers.
The following are the main reasons behind the adoption of Ethernet in carrier networks:
Ethernet is everywhere. Retail as well as corporate customers are already using
Ethernet. By adopting Ethernet in the backbone, the same technology can be used
from end-to-end, thus eliminating the need for interface or frame conversion.
Cost effectiveness. Due to economies of scale, Ethernet components are inexpensive,
thus lowering both the capital and operational expenses.
Increased flexibility. Unlike traditional point-to-point circuits that come in discrete
speeds (T1/E1, T3/E3, OC-3/STM-1), Ethernet allows subscribers to add bandwidth
more incrementally and more quickly. Typically, bandwidth upgrades only involve
configuration changes and can take effect in days rather than weeks.
To promote the use of carrier Ethernet and to increase interoperability between
vendors, the Metro Ethernet Forum (MEF), an industry consortium, produces technical
specifications for Ethernet services in metropolitan areas as well as in the wide area.
In addition to extending Ethernet service to the customers, carriers and service
providers are also building and converting their own core networks to Ethernet.
For example, more and more mobile phone operators and DSL providers are
using Ethernet as the back-haul technology instead of the traditional TDM or
ATM network[1].
QoS
Over the years, Ethernet has evolved from its original best-effort, CSMA/CD based
data communications technology to a QoS (Quality of Service) enabled service
capable of supporting multi-media traffic in a converged network. As an example,
MEF uses the concept of a bandwidth profile to describe the subscriber’s traffic.
It specifies the average rate of "committed" and "excess" traffic that a subscriber can
generate into the provider’s network, and the associated performance objectives in terms
of the delay, frame loss and availability. Thus each traffic profile consists of the following
parameters:
- CIR (Committed Information Rate)
- CBS (Committed Burst Size)
- EIR (Excess Information Rate)
- EIS (Excess Information Size)
Service frames up to the committed rate are considered in-profile and are delivered
as per the performance objectives. Those sent up to the excess rate are allowed into the
provider’s network but are considered out-of-profile and are delivered without any service
performance objectives. Traffic sent above the excess rate is discarded. Thus subscribers
who have different bandwidth profiles will be accorded a different level of service. [2]
Field Testing Requirements
In an increasingly competitive environment, carriers and service providers must be able to
provision circuits quickly and cost-effectively, while at the same time ensuring that the
service meets or exceeds the Service Level Agreement (SLA). This requires engineers and
technicians to be able to accurately and efficiently test circuits in the field. Field testing
generally falls into a few categories – basic connectivity testing and service verification;
performance testing and QoS testing.
To facilitate field testing, portability of the test equipment becomes an important
consideration. Thus instead of chassis based units, hands-held devices are commonly
used. In recent years, these units have become widely available and affordable.
The SunLite GigE from Sunrise Telecom is one such example and will be used to illustrate
the tests described in this white paper.
Basic Connectivity Testing and Service Verification
Whether it is a new customer or an existing customer ordering a new circuit, the field
technician must ensure that the circuit is operational before it is turned-up. This type
of testing usually consists of the following:
- Link check – confirm the link is active and successfully negotiated a speed and duplex setting
- DHCP – confirm that the CPE can obtain a valid IP address
- PING – confirm that there is IP connectivity
- Traceroute – confirm that IP traffic is taking the desired route
These tests are also commonly used to troubleshoot an Ethernet network. (See sidebar on
Ethernet Troubleshooting) Additional tests such as loading a web page using HTTP or
downloading a file using FTP are also often performed in order to verify that Internet
connectivity is established.
Performance Testing
Basic connectivity testing is generally sufficient for best-effort service such as residential
Internet access which has no implicit performance guarantees. For corporate customers
who require services with specific performance objectives, it is common to employ the
RFC 2544 tests.
RFC 2544 [3], published in 1999 by the IETF, defines the “Benchmarking Methodology for
Network Interconnect Devices”. It was originally designed to allow the standardized testing
and benchmarking of a single interconnect device such as a router or a switch (known as
the DUT or Device Under Test). This methodology has become the de facto standard
performed routinely in QA labs and verification labs in order to quantify the performance
of network devices. The following diagram shows the classical test setup.

In this setup, a test system is responsible for injecting traffic into the DUT and then
analyzing the frames coming out. By correlating the input and output, the test system can
measure the transmission characteristics of the DUT.
The RFC 2544 benchmarking methodology defines a series of performance measurements.
The terminology and metrics used in the test suite is defined in [4]. The following are the
four benchmark tests that are most commonly implemented in an automated test suite1:
Throughput – This test measures the highest data rate at which none of the frames are
dropped by the DUT. The following frame sizes in bytes should be used during the test – 64,
128, 256, 512, 1024, 1280 and 1518.
Latency – Once the throughput has been determined, this test measures the one-way delay
through the DUT at the throughput rate.
Frame loss – This test measures the frame loss rate throughout the entire range of input
data rates and frame sizes.
| FrameLoss | | = | (InputCount - OutputCount)×100%
 InputCount |
|
|
Back-to-back – This test measures the longest frames burst (frames transmitted with
the minimum inter-frame gaps in between) that the DUT can handle without the loss of
any frames.
| 1In addition to these four, RFC 2544 also defines the system recovery test and the reset test which measure the
speed at which a DUT recovers from an overload condition and a device reset. These are not used as much mainly
because these metrics are typically not part of the carrier SLA. Furthermore, these two tests are also harder to
implement in an automated test. |
In recent years, the RFC 2544 test suite has been adapted to measure the performance of
systems and networks. Thus instead of measuring the ability of a switch or router to pass
frames, we can now use the same methodology to quantify the transmission characteristics
of a virtual circuit or a network. The following shows two examples. The first one shows an
example of what MEF defines as the Ethernet Line (E-Line) service, which is essentially an
emulation of a virtual circuit. The second example is an Ethernet LAN (E-LAN) service, which
emulates a multipoint network.


Adapting RFC 2544 from testing a DUT in the lab to a production network in the field
requires a few changes:
- The test system used in the test lab connects ports that are physically close
to each other. Typically, the same test system transmits traffic from one
set of ports and receives traffic from another set of ports. In a network,
the two measurement points are geographically separated. Thus two or
more test units are used to measure the input and the output. In other
words, one unit generates the traffic and the other unit receives the traffic.
- In the lab, the test system is stationary. In the field, the test unit must be
portable and can be deployed quickly in various locations.
- RFCs 2544 and 1242 define latency in terms of the one-way delay. In a
network where two or more test units are required to perform the test,
round-trip delays are measured. Measuring one-way delay would require the
test units to be time-synchronized.
The following are some sample results of a RFC2544 throughput test of a Gigabit switch:
TEST MODE : THROUGHPUT TEST
SPEED MODE : 1000 FULL
FLOWCONTROL : OFF
FLOWCONTROL : OFF
PREAMBLE SIZE : 8
DATA PATTERN : RANDOM
TX RATE(INIT) : 60%
TX RATE(FINAL) : 100%
TX RATE(DELTA) : 10%
TX PACKETS : 10000
64 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
128 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
256 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
| |
512 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
768 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
1024 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
1280 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
1518 BYTES :
MAX RATE : 100%
FRAME LOSS : 0
PASS/FAIL : PASS
|
The automated test stepped through all the prescribed frame sizes testing each at the line rate.
In this test, the switch was able to pass through wire rate traffic for all frame sizes.
Asymmetric Testing
In field testing, another common test scenario is the broadband access network with
asymmetric bandwidth. In ADSL, for example, it is important to test the two directions
(upstream and downstream) separately. This is illustrated in the following diagram.
Furthermore, by inserting the test units at various points of the upstream network, we can
determine the throughput of the regional as well as access networks.

The following shows the two test configurations for measuring the downstream and
upstream performance.

Field Testing of QoS Enabled Network
Today’s multi-service networks must support a variety of traffic such as data, voice
and video. Corporate clients also have SLAs with stringent performance requirements.
For example, some carriers offer a gold, silver and bronze service, each with a different
set of service objectives. Thus increasingly, the network infrastructure must be able
to differentiate and prioritize traffic. This is usually done using one or more of the
following mechanisms.
Physical Port. Traffic can be given different priorities simply based on the originating port.
In a private network, the physical port may correspond to, say, the voice VLAN where all the
IP phones are. In a carrier network, each port may correspond to a different client with a
different SLA.
802.1p. Within a switched network, Ethernet frames can be tagged with an additional
802.1p header.2 This provides QoS at the MAC level. There are three priority bits (referred
to as Priority Code Point or PCP) which indicate the priority of the frame. This is sometimes
referred to the Class of Service (CoS).
IP TOS or Diffserv. The network can also provide QoS at the IP layer using the TOS (Type of
Service) or Diffserv (Differentiated Services). Unlike the 802.1p bits which are usually set by
the end system, the TOS and Diffserv bits are usually assigned by the router based on either
the IP address of the source or the type of traffic (TCP or UDP port numbers).
VLAN. In an aggregation network, all the subscribers within the same service class can be
assigned to a unique VLAN. For example, in a DSL network, a DSLAM can assign all the Gold
service subscribers to the same VLAN. The backbone network can then prioritize traffic based
on the VLAN ID. This VLAN tag is usually referred to as the S-VID or the provider VLAN. In
order to preserve the customer’s own VLAN designation (C-VID), Q-in-Q is often used.
Testing Prioritization
Regardless of the prioritization scheme, one must then be able to confirm that indeed the
network is prioritizing traffic in accordance with the service level objectives. Testing QoS
requires the ability to inject streams of traffic that will be accorded different treatment by
the network and then measure their respective performance. The following diagrams
illustrate a few scenarios.

2The 802.1p header is an 802.1Q header with the VLAN ID set to 0.
In this test setup, 3 streams of traffic are injected into the network each with a different
PCP (0 being the lowest and 7 the highest priority). On the receiving side, the test set
monitors each stream to look for impairments such as packet loss. In order to verify that
traffic is being prioritized, there must be congestion on the output port, thus forcing the
network to decide which packets to drop. This congestion can be simulated if necessary by
injecting some additional background traffic from another port.3

The next two scenarios illustrate similar setups for verifying IP prioritization as well as
prioritization based on VLAN.

3These few diagrams show a separate port and a separate test unit used to generate the background traffic. This
is only necessary if the ingress and egress ports have the same speed and thus additional traffic is needed to
artificially create a congestion condition.

The last example shows an example of Q-in-Q. Each test packet has two 802.1Q tags, the
outer tag is the provider tag and the inner tag is the customer tag. In this example, a
provider tag of 100 might indicate a “Gold” service and 101 might indicate a “Silver” service.
Thus the test will determine if each provider tag indeed accords each class with a different
service level.
Using the SunLite GigE, one can construct multiple simultaneous packet streams and
perform a “BERT Throughput” test. In the example below, four streams are enabled that
correspond to four different bandwidth profiles. Each packet is configured with 2 VLAN tags
(Q-in-Q). The outer tag (S-VID) has a VLAN ID of 100 whereas the inner tag (C-VID) has a
VLAN ID of 200.
|
Troubleshooting Ethernet
What to do when the network does
not work? Here is where troubleshooting
tools come in. In Ethernet,
there are a number of common
problems that can be identified
quickly with the right tools. These
include the following:
- Cabling error. The LAN cable is
connected to the wrong port.
- Cable fault. There is a problem with
the connector resulting in an open
or short circuit.
- Speed or duplex mismatch. The NIC
is configured for a speed/duplex
setting that is incompatible with
that of the switch port.
- DHCP error. Stations are unable
to obtain an IP address from the
DHCP server.
- Duplicate IP address. Multiple
devices are configured statically
with the same IP address.
- Routing problem. Stations are
unable to reach destinations
outside the local VLAN either
because the default gateway is
down or misconfigured.
To diagnose these problems quickly, the field personnel needs a number of troubleshooting tools that are now packaged in hand-held test units.
The following are some of these common tools:
- Cable tester – This function checks for any cable faults such as open and short circuits. Using TDR (Time Domain Reflectometry), the tester can also
determine the length of the connected cable.
- Port identifier – This function activates and deactivates the Ethernet port thus causing the port LED on the switch to turn on and off periodically.
This is useful in identifying the switch port connected to a device.
- Link status check – When the test unit is connected to a switch port, this function displays the link status such as the speed, duplex and flow control.
- DHCP – This allows a test unit to confirm that an IP address can be acquired successfully from the DHCP server.
- Device scan – This function scans a range of IP addresses looking for active devices.
- PING – This standard function probes an IP address using ICMP. It returns the round-trip delay between the test unit and the target device.
- Traceroute – This standard function traces the IP path to a target address identifying all the intermediate routers.
|
|
After configuring the streams, traffic was turned on. The following screens show the test results:
Jitter or Packet Delay Variation
Real-time traffic such as IP telephony or IP video is sensitive not only to the amount of
delay but also whether the amount of delay is relatively constant or varying. This measure
is referred to as the packet jitter or packet delay variation4. To complicate matters, there are
many ways of calculating delay variations and they can yield different results. The following
are a few commonly used definitions:
- In RFC 3393, the one-way IPDV (IP delay variation) is defined as the difference
between the delays experienced by two consecutive packets from the same stream.5
- In RFC 4689, jitter is defined as the absolute value of the IPDV.6
- MEF defines frame jitter as the difference between the maximum frame delay
and the minimum frame delay within a group of samples.7
- In ITU Y.1540, the 2-point Packet Delay Variation is defined as the difference
between the packet transfer delay of the packet and a defined reference IP
packet transfer delay.8
Triple Play
For real-time applications such as voice and video, network transmission attributes such
as packet delay, jitter and packet loss can adversely affect the quality of the reception. For
example, for IP telephony, the following are commonly recommended values in order to
maintain acceptable voice quality:
- Packet delay: Less than 150 ms one-way (less than 100 ms one-way for tollquality voice).
- Jitter: Less than 50 ms
- Packet loss: Less than 1%
For triple play, one must be able to generate multiple data streams and verify that voice
and video traffic is transmitted with the required performance. In this test, the different
streams can represent data, voice and video. How the network treats each stream depends
on the network configuration. In some networks, all IP phones are assigned to a separate
voice VLAN. In others, the IP phones are programmed to tag all traffic with higher priority
using 802.1p. Thus in designing a test, one must take into consideration the configuration
of the network infrastructure in order to effectively simulate the traffic mix. The following
diagram illustrates a typical setup.
4Packet jitter or delay variation measures the change in packet delay. This is not to be confused with video jitter
which happens when the picture “jumps”. Video jitter can be the result of packet loss and excessive packet jitter.
Future Requirements - Ethernet Service OAM
As carrier Ethernet becomes more widely deployed, an area which is evolving rapidly is Ethernet
Service OAM (Operation, Administration and Maintenance). The objective is to provide
carriers with a comprehensive set of tools to monitor the availability and performance of
Ethernet services. There are a number of related efforts in progress:
- IEEE 802.3-2005 (previously 802.3ah) – Ethernet Link OAM (Ethernet in
the First Mile)
- IEEE 802.1ag – Ethernet Service OAM
- ITU-T Y.1731 – OAM Functions and Mechanisms for Ethernet Based Networks
- MEF 17 – Service OAM Requirements and Framework
It is expected that as these standards mature, the prescribed tests (such as connectivity
check, loopback, link trace and so on) will be implemented in the field test sets.
Conclusions
Ethernet has become the ubiquitous networking technology. Carriers and service
providers are under competitive pressure to provision service quickly and cost-effectively.
One of the challenges of the field technician is to verify the integrity of the service which
is becoming more complex. Basic testing uses simple IP tools such as DHCP, PING and
traceroute to verify connectivity. Increasingly, Ethernet service must also be measured
against the service level. RFC 2544 is the test suite that can be used to determine the
throughput, latency, frame loss and back-to-back. In a QoS enabled network, traffic may
be prioritized depending on the physical port, 802.1p, TOS or DiffServ, or VLAN tag. To verify
proper QoS handling, the field engineer must be able to generate multiple packet streams
with different characteristics and measure the performance of each stream separately.
For example, in a network that uses Q-in-Q, the packets must be configured with at least
two VLAN headers to allow the aggregation switches to prioritize traffic using the outer tag.
Finally, to ensure that the network can handle real-time traffic properly, the packet delay
variation (which is not part of the RFC 2544 test) must also be included.
References
1 TR-101, “Migration to Ethernet Based DSL Aggregation”, Broadband Forum, April 2006.
2 MEF 10.1, “Ethernet Services Attributes Phase 2”, Metro Ethernet Forum, November 2006.
3 Bradner, S. and McQuaid, J., “RFC 2544 – Benchmarking Methodology for Network Interconnect
Devices”, IETF, March 1999.
4 Bradner, S., “RFC 1242 – Benchmarking Terminology for Network Interconnection Devices”,
IETF, July 1991.
5 Demichelis, C. and Chimento, P., “RFC 3393 - IP Packet Delay Variation Metric for IP Performance
Metrics (IPPM)”, IETF, November 2002.
6 Poretsky, S., Perser, J., Erramilli, S., and Khurana, S., “RFC 4689 – Terminology for Benchmarking
Network-layer Traffic Control Mechanisms”, IETF, October 2006.
7 MEF 6.1, “Ethernet Services Definition – Phase 2”, MEF, April 2008.
8 ITU-T Y.1540 “Internet Protocol Data Communication Service – IP Packet Transfer and
Availability Performance Parameters”, December 2002.