Understanding Carrier Ethernet Service Assurance

Taking advantage of MEF ’s best practices for performance monitoring

— Part II —
How to Measure Carrier Ethernet Performance Attributes

September 2016

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 1 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

Table of Contents
1. Introduction and Overview __________________________________________________ 4
1.1 Abstract____________________________________________________________________ 4
1.2 Target audience _____________________________________________________________ 4
1.3 Document Purpose and Scope __________________________________________________ 4
1.4 Executive Summary __________________________________________________________ 4
2. Carrier Ethernet Service Performance _________________________________________ 5
2.1 Performance Attributes – How to Measure _______________________________________ 5
2.1.1 Service OAM Overview ______________________________________________________________ 5
2.1.2 MEF Service OAM and ITU-T G.8013/Y.1731 _____________________________________________6
2.1.3 Performance Monitoring (PM) Measurements ___________________________________________6
2.1.4 One-way and two-way measurements versus single-ended and dual-ended messaging ___________7
2.1.5 Performance Monitoring (PM) Functions and Solutions ____________________________________8
2.1.6 PM Sessions and Measurement Intervals _______________________________________________12
2.1.7 Proactive versus on-demand monitoring _______________________________________________12
2.1.8 Required instrumentation __________________________________________________________ 13
2.1.9 Threshold Crossing Alerts (TCAs) _____________________________________________________ 14
2.1.10 Performance Objectives and CoS ___________________________________________________ 14
2.1.11 Practical Considerations and Terminology ____________________________________________15
3. Summary _______________________________________________________________ 17
4. Guide to the MEF Technical Specifications ____________________________________ 18
5. About the MEF __________________________________________________________ 20
6. Glossary and Terms _______________________________________________________ 20
7. References ______________________________________________________________ 20
8. Acknowledgements _______________________________________________________ 20

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 2 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

List of Figures
Figure 1 – Diagram of Example MEG Levels............................................................................................................ 6
Figure 2 – Single-ended Synthetic Loss Measurement.......................................................................................... 10

List of Tables
Table 1 - PM Solutions Summary ........................................................................................................................... 7
Table 2 - Performance Attributes and Relevant PM Functions ............................................................................... 9
Table 3 - PM-1 Solution’s Supported Performance Attributes .............................................................................. 10
Table 4 - PM-2 Solution’s Supported Performance Attributes .............................................................................. 11
Table 5 - PM-3 Solution’s Supported Performance Attributes .............................................................................. 11
Table 6 - PM-4 Solution’s Supported Performance Attributes .............................................................................. 12
Table 7 - Performance Terminology ..................................................................................................................... 15

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 3 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

1. Introduction and Overview

1.1 Abstract
The MEF has addressed multiple aspects of service assurance throughout its technical specifications to
enable the necessary performance and service quality available with Carrier Ethernet 2.0 (CE 2.0)
services.

The MEF’s Service Management Life Cycle white paper describes the different stages in the life cycle of a
CE 2.0 based service including performance monitoring and fault management. This paper together with
its companion paper, “Part I: An Introduction to Service Assurance and Carrier Ethernet Service
Performance Attributes”, provides an in-depth guide to performance monitoring and assurance across
the body of work of the MEF enabling the reader an accessible, one-stop guide to understanding CE 2.0
performance monitoring best practices and use them to their full advantage.

1.2 Target audience

Communication Service Providers (CSPs) including telecom service providers, Internet service provider
(ISP), cable operators/MSOs, cloud service providers, wireless network operators, network equipment
providers, and OSS/LSO providers.

1.3 Document Purpose and Scope

This paper explores the details of how to measure and monitor performance in the context of CE 2.0
services, the Third Network vision and Lifecycle Service Orchestration (LSO).

Those familiar with some of the key technical specifications such as MEF 35.1 (Service OAM
Performance Monitoring Implementation Agreement) as well as those unfamiliar with the work of the
MEF or its many technical specifications will benefit from this holistic view covering the current MEF
body of work with respect to the critical and complex topic of service assurance.

1.4 Executive Summary

Service assurance encompasses the procedures used to monitor a network service to maintain a
particular service quality. This paper focuses on the practical considerations and methods for measuring
Carrier Ethernet service performance. The reader is encouraged to read the companion paper, “Part I:
An Introduction to Service Assurance and Carrier Ethernet Service Performance Attributes”, where the
service performance attributes are defined.

Based on the relevant business drivers and performance attributes, the reader will benefit from
understanding how to measure the service quality of Carrier Ethernet services based on MEF standards.
There exists significant confusion and misconceptions with respect to performance management best
practices for Carrier Ethernet services. This confusion is the result of multiple transport technologies,
legacy tools, proprietary instrumentation, interoperability challenges as well as other related service
assurance standards making it difficult to understand and leverage the best practices available to Carrier
Ethernet service providers. By clarifying the what and the how, the industry can more quickly align to
simplify management, accelerate delivery and turn-up of existing services as well as new service
launches which together improve the critical time-to-revenue factor for providers.

Having the performance-oriented service assurance elements which are specified in over a dozen
technical specifications centralized and accessible to the reader enables a more efficient and rapid
MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 4 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
implementation of the industry best practices for Carrier Ethernet 2.0 service assurance from the
definition of service quality objectives, to the proper monitoring, presentation and sharing of the
resulting service performance intelligence.

2. Carrier Ethernet Service Performance

The companion to this paper, “Part I: An Introduction to Service Assurance and Carrier Ethernet Service
Performance Attributes”, examined the defined Carrier Ethernet Service Performance Attributes. This
paper defines how those attributes are measured.

2.1 Performance Attributes – How to Measure

2.1.1 Service OAM Overview
The Service OAM (operations administration and maintenance) framework includes the fault
management, performance monitoring, auto-discovery and both intra- and inter-provider service OAM.
It focuses on the Ethernet Services layer (ETH Layer) as opposed to the transport service layer (e.g. MPLS
or SONET/SDH) or the Application Services Layer (e.g. IP, not to be confused with the subscriber’s
definition of the application such as an enterprise resource planning application).
Service OAM works in conjunction with both the control plane and management plane in an effort to
measure performance and to monitor the data plane that carries traffic. The Service OAM framework
defines the points in the network that must be maintained and groups them into Maintenance Entity
Groups (MEGs). These groups contain end points (MEPs) and intermediate points (MIPs). Performance
monitoring takes place at the MEP. Furthermore, it defines the protocol data units (PDUs) used for
SOAM traffic and how that is handled through the CEN.
To handle the different administrative layers of a network service (e.g. the last mile connection, an
access provider segment in a multi-operator service, the service provider segment and the subscriber
segment), MEGs are defined with a level of 0 to 7 so that lower level MEGs allow SOAM traffic from
higher level MEGs to pass through unaffected. For example, the customer domain MEG below would be
defined as MEG level 6. The operator domains (in orange) could be MEG level 2 and that level 2 domain

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 5 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
would pass the MEG level 6 SOAM traffic through unaffected so that metrics for the customer UNI to
UNI performance attributes could be measured.

Figure 1 – Diagram of Example MEG Levels

2.1.2 MEF Service OAM and ITU-T G.8013/Y.1731

To help provide guidance on how to best implement the SOAM performance monitoring for CE services,
the MEF created an implementation agreement (MEF 35.1). The ITU-T Recommendation
G.8013/Y.17311 - OAM Functions and mechanisms for Ethernet-based networks, (colloquially referred
to as “Y.1731”) provides the instrumentation to calculate each of the performance metrics that make up
the SLS. Some confusion exists in the marketplace because of the fact that Y.1731 is partially
implemented by many equipment vendors leading some operators to believe they are using MEF-
defined performance monitoring, however though the MEF takes advantage of the various PM functions
within Y.1731, the actual performance metrics and requirements for collection, measurement and
storage are uniquely defined in MEF 35.1.

2.1.3 Performance Monitoring (PM) Measurements

The MEP (Maintenance Entity Group End Point) of an associated SOAM MEG level is able to both
generate and respond to SOAM PDUs. The MIP (Maintenance Entity Group Intermediate Point) can only
respond and is not used for performance monitoring functions.

The first key concept is that all measurements are performed at the network element layer (i.e. within
the network equipment). The measurements, known officially as PM measurements may then be sent
northbound to the Element management layer to an EMS (element management system) and in turn to
the network management layer to an NMS (network management system). Alternatively, the PM
measurements can be sent directly to the NMS (or queried/polled by the NMS). Given the recent work

1
As of July 2011, the ITU-T Y.1731 recommendation was renamed G.8013/Y.1731; https://www.itu.int/rec/T-REC-
Y.1731
MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 6 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
of the MEF on LSO, the legacy terms EMS and NMS are beginning to transition to Element Control and
Management (ECM) and Infrastructure Control and Management (ICM) respectively.

The MEF defines four performance monitoring (PM) solutions which refer to particular MEG types
(point-to-point or multi-point), measurement techniques for loss measurements and the PM function(s)
defined to generate the appropriate metrics for the solution. These solutions are outlined in the table
below, where PM Solution 1 (PM-1) is mandatory, the others are optional.

Table 1 - PM Solutions Summary

Measurement Mandatory
PM MEG Technique or
Solution Type(s) for Loss PM Function(s) Optional
PM-1 point-to-point Synthetic Single-Ended Delay Mandatory
multipoint Testing Single-Ended Synthetic Loss
PM-2 point-to-point n/a Dual-Ended Delay Optional
multipoint
PM-3 point-to-point Counting Service Single-Ended Service Loss Optional
Frames
PM-4 point-to-point Synthetic Dual-Ended Synthetic Loss Optional
multipoint Testing

The PM-1 mandatory solution applies to all EVC (and OVC) service types and uses synthetic tests to
measure loss. Synthetic testing uses test traffic to measure loss versus actually measuring real frame
loss between two UNIs. Synthetic testing can provide visibility on network performance even when the
network is not in use, helping establish early signs of performance degradation. In addition, the
synthetic test functions are generally simpler computationally, however they add traffic to the network
and ultimately do not exactly reflect real frame loss for qualified service frames.

2.1.4 One-way and two-way measurements versus single-ended and dual-ended messaging
The familiar TCP/IP tool: ICMP Ping enables a user to initiate a "Ping" test from a source IP address to a
destination IP address and the command returns the round trip time for the ICMP packet to traverse
from the source to the destination and back again. This common example may help in visualizing the
terminology used throughout the MEF specifications with respect to the PM tools (i.e. those borrowed
from Y.1731) and the PM Functions described in section 2.1.5.

2.1.4.1 One-way vs. Two-way Measurements

There is a fundamental difference between a one-way measurement and a two-way measurement.
Given two end points (A and Z), a one-way measurement will measure the value (e.g. delay) going from
one end point to the other. For example, the one-way delay between A and Z can be measured from A
to Z, and separately from Z to A. A one-way delay measurement for instance, is very useful as opposed
to the round trip time common to a "Ping" test, given that the A-to-Z value is measured distinctly from
the Z-to-A value which may demonstrate asynchronous results. In fact, the ITU-T defines one-way in the
context of Ethernet frames in G.8001 as a unidirectional measurement. All of the MEF performance
attributes are defined as one-way.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 7 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
Conversely a two-way measurement is analogous to the "ping" test and is defined once again by ITU-T as
a bidirectional or round-trip measurement. There are no round-trip-based performance attributes
within the MEF specifications, however the round-trip frame delay can be used to estimate the one-way
delay by assuming synchronous delay and dividing the round-trip result by two to approximate the one-
way delay measurement. However, to achieve an accurate one-way measurement for delay, both the
source (controller MEP) and destination (sink MEP) must have their clocks synchronized. Clock
synchronization is not required for measuring one-way inter frame delay variation (IFDV) nor one-way
frame delay range (FDR).

2.1.4.2 Single-ended vs. Dual-ended Messaging

ITU-T uses the terms single-ended and dual-ended for defining their PM tools. These terms differ slightly
from the terms one-way and two-way measurements. The primary difference is that a measurement is a
specific test result from the PM tool(s), that is, one-way delay measures the delay in one-direction from
a source to a destination. The terms single-ended and dual-ended are not used to describe the
measurement but rather how the measurement is achieved, specifically the messaging protocol.

For example, the ITU-T defines single-ended delay as a tool with a controller MEP that sends a particular
message to a responder MEP that in turn replies with a related message back to the controller. This
sounds a lot like the "Ping" test, however it isn’t directly related to a measurement, it merely indicates
that one MEP, the controller MEP, initiates the tool (sends the initial PDUs) and performs the
calculations once the test is complete. The responder MEP doesn’t get involved in any calculation in this
instance, hence it is referred to as a ‘single-ended’ test. That test can provide a two-way measurement
as well as a one-way measurement, however for the one-way measurement to be accurate in the case
of a delay measurement, the two MEPs must have their clocks synchronized.

Conversely a dual-ended test requires that the controller MEP (initiating) sends the request PDU to the
sink MEP (destination) and the sink MEP performs all of the calculations. There is no response or reply
back to the controller MEP. In this case only a one-way measurement can be calculated, although
academically speaking, a round-trip delay time could be estimated by doubling the one-way
measurement but this is not done in practice, nor is it relevant to the MEF performance attributes which
are all defined as one-way.

Typically, the choice to use single-ended versus dual-ended is based on a CSP’s operational preferences.
For example, dual-ended messaging means that the PM tools can be configured symmetrically on all
devices. Furthermore, in multi-point networks, the test PDUs can be multicast to make more efficient
use of the network. Otherwise, with the exception that dual-ended tests cannot provide two-way
measurements, single-ended and dual-ended messaging are equivalent in terms of the
measurements they can provide.

Note: The terms sink MEP and responder MEP are similar in that the controller MEP (initiating the
tests) sends the test PDUs to the sink/responder, however there is a distinction in that a sink MEP is
exclusive to dual-ended tests and there is no reply/response. A responder MEP on the other hand is
exclusive to single-ended tests and responds to the controller MEP.

2.1.5 Performance Monitoring (PM) Functions and Solutions

There are 4 PM Solutions defined which make use of five PM functions, single-ended delay, single-ended
synthetic loss, dual-ended delay, single-ended service loss and dual-ended synthetic loss. PM-1 Solution
is mandatory. PM-2, PM-3 and PM-4 Solutions are optional.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 8 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

The relationship between each performance attribute and the associated PM functions which can
provide measurements for the attribute is illustrated in the table below.

Table 2 - Performance Attributes and Relevant PM Functions

Performance Attribute PM Function

One-way Frame Delay Performance Single-ended delay (PM-1 Solution)
Dual-ended delay (PM-2 Solution)
One-way Mean Frame Delay Performance Single-ended delay (PM-1 Solution)
Dual-ended delay (PM-2 Solution)
One-way Frame Delay Range Performance Single-ended delay (PM-1 Solution)
Dual-ended delay (PM-2 Solution)
One-way Inter-Frame Delay Variation Single-ended delay (PM-1 Solution)
Performance Dual-ended delay (PM-2 Solution)
One-way Frame Loss Ratio Performance Single-ended synthetic loss (PM-1 Solution)
Single-ended service loss (PM-3 Solution, point-to-point
EVC/OVC only)
Dual-ended synthetic loss (PM-4 Solution)
One-way Availability Performance Single-ended synthetic loss (PM-1 Solution)
Dual-ended synthetic loss (PM-4 Solution)
One-way Resiliency Performance Single-ended synthetic loss (PM-1 Solution)
expressed as High Loss Intervals Dual-ended synthetic loss (PM-4 Solution)
One-way Resiliency Performance Single-ended synthetic loss (PM-1 Solution)
expressed as Consecutive High Loss Dual-ended synthetic loss (PM-4 Solution)
Intervals
One-way Group Availability Performance Single-ended synthetic loss (PM-1 Solution)
Dual-ended synthetic loss (PM-4 Solution)
Composite Performance Metric (CPM) Either single-ended delay or dual-ended delay coupled
with either single-ended synthetic loss, dual-ended
synthetic loss or in the case of point-to-point EVC/OVC,
single-ended service loss

The PM functions are described below, organized by their relevant PM solution.

2.1.5.1 PM-1 Solution (for point-to-point and multipoint services)

Single-ended delay
This is based on initiating a test from a controller MEP to a responder MEP using ITU-T’s single-ended
ETH-DM tool defined in Y.1731. The PDUs used are DMM (delay measurement message) and the DMR
(delay measurement response).

Note: Single-ended delay tests need clock synchronization between the controller and responder
MEPs when used to provide accurate one-way Frame Delay Performance measurements for the
purpose of calculating FD and MFD performance metrics. Otherwise, depending on the discrepancy
in time-of-day between the two MEPS, the FD (and MFD) metrics will be suspect. Alternatively, in
the absence of clock synchronization, these tests can provide two-way Frame Delay Performance
measurements which can be used to estimate the one-way Frame Delay Performance.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 9 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
Single-ended synthetic loss
This is based on initiating a test from a controller MEP to a responder MEP using ITU-T’s single-ended
ETH-SLM tool defined in Y.1731. The PDUs used are SLM (synthetic loss message) and SLR (synthetic loss
response). Synthetic loss is easily defined but may not represent the actual behavior experienced by real
qualified service frames. In addition, a statistically significant quantity of tests is necessary to provide a
reasonable estimate of frame loss.

Figure 2 – Single-ended Synthetic Loss Measurement

Single-ended synthetic loss measurement example

In this example an SLM message is created at the controller MEP (MEP-A) which includes the CoS ID, and
the local counter value containing the number of SLM messages sent. The SLR message is generated in
response at the responder MEP (MEP-B) which includes the count of received messages. MEP-A then
calculate the one-way frame loss performance based on the data in the SLR message.

Table 3 - PM-1 Solution’s Supported Performance Attributes

Supported Performance Attribute

One-way Frame Delay Performance
One-way Mean Frame Delay Performance
One-way Frame Delay Range Performance
One-way Inter-Frame Delay Variation Performance
PM-1 Solution One-way Frame Loss Ratio Performance
One-way Availability Performance
One-way Resiliency Performance expressed as High Loss Intervals
One-way Resiliency Performance expressed as Consecutive High Loss
Intervals
One-way Group Availability Performance

2.1.5.2 PM-2 Solution (for point-to-point and multipoint services)

Dual-ended delay
This is based on initiating a test from a controller MEP to a sink MEP using ITU-T’s dual-ended ETH-DM
tool defined in Y.1731. The PDU used is 1DM (single delay measurement).

Note: Dual-ended delay tests need clock synchronization between the controller and
sink MEPs when used to provide accurate one-way Frame Delay Performance
measurements for the purpose of calculating FD and MFD performance metrics .
Otherwise, depending on the discrepancy in time -of-day between the two MEPS, the

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 10 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
FD (and MFD) metrics will be suspect. Dual-ended delay tests cannot provide two -way
Frame Delay measurements.

This PM solution is optional and applies to both point-to-point and multi-point services. It does not
include any loss metrics, but uses the dual-ended delay PM function providing measurements for FD,
MFD and FDR as well as IFDV based on the 1DM PDU. Note that 1DM PDUs can be multicast for efficient
performance monitoring of multi-point services such that a single controller MEP can send the PDU to
multiple sink MEPs.

Table 4 - PM-2 Solution’s Supported Performance Attributes

Supported Performance Attribute

One-way Frame Delay Performance
PM-2 Solution One-way Frame Delay Range Performance
One-way Mean Frame Delay Performance
One-way Inter-Frame Delay Variation Performance

2.1.5.3 PM-3 Solution (for point-to-point services)

Single-ended service loss
This is based on initiating a test from a controller MEP to a responder MEP using ITU-T’s single-ended
ETH-LM tool defined in Y.1731. The PDUs used are LMM (loss measurement message) and LMR (loss
measurement response). This tool can provide the most accurate measurement of real frame loss
between 2 UNIs as it is derived from real subscriber traffic. However, during low periods of usage,
aggregate metrics (averages, peaks, etc.) may not reflect the actual performance of the service as
compared to synthetic measurements with a statistically significant amount of test traffic. This solution
can only be applied to a point-to-point EVC (or OVC) since all ingress service frames are destined for the
other UNI as opposed to a multipoint service where ingress service frames may be destined for different
UNIs.

This PM solution is optional and applies only to point-to-point services given that it counts the actual
service frames that egress from the responder MEP and compares this to the number of frames sent at
ingress to the controller MEP.

Table 5 - PM-3 Solution’s Supported Performance Attributes

Supported Performance Attribute (Point-to-point EVC/OVC only)

PM-3 Solution
One-way Frame Loss Ratio Performance 2

2
FLR is used in the calculation for resiliency (HLI, CHLI) and Availability, thus the reader may question
why these attributes are not supported by PM-3 Solution along with FLR. The primary reason is that the
One-way Availability Performance attribute is meant to provide continuous, proactive monitoring. Because
PM-3 solution requires traffic on the network to take a measurement, there is no way to measure
availability if the service is idle (no traffic), therefore PM-3 can only be used for one-way Frame Loss
Ratio performance.
MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 11 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
2.1.5.4 PM-4 Solution (for point-to-point and multipoint services)
Dual-ended synthetic loss
This is based on initiating a test from a controller MEP to a sink MEP using ITU-T’s single-ended ETH-SLM
tool defined in Y.1731. The PDU used is 1SL (single synthetic loss). Synthetic loss is easily defined but
may not represent the actual behavior experienced by real qualified service frames. In addition, a
statistically significant quantity of tests is necessary to have a reasonable estimate of frame loss.

This PM solution is optional and applies to both point-to-point and multipoint services. It is based strictly
on dual-ended synthetic loss measurements, that is, true one-way measurements from a controller MEP
to one or more sink MEPs. The 1SL PDUs can be unicast or multicast for efficient performance
monitoring of multipoint services.

Table 6 - PM-4 Solution’s Supported Performance Attributes

Supported Performance Attribute

One-way Frame Loss Ratio Performance
One-way Availability Performance
PM-4 Solution One-way Resiliency Performance expressed as High Loss Intervals
One-way Resiliency Performance expressed as Consecutive High Loss
Intervals
One-way Group Availability Performance

2.1.6 PM Sessions and Measurement Intervals

A performance monitoring session or PM session is a term used to describe the specific PM solution
(and its PM function), such as single-ended delay between two MEPs for a specific CoS. Performance
measurements are partitioned into fixed-length Measurement Intervals. The SLS objectives are defined
over a period of time T. Typically T is made up of multiple measurement intervals.

The MEF does not dictate the specific measurement intervals that should be used by the industry. Such
decisions must be made by each CPS, based on the specific service, its underlying transport technology
and available instrumentation as well as the market conditions that govern expected or common SLAs
which subscribers expect.

2.1.7 Proactive versus on-demand monitoring

These PM tools can be initiated through a PM session on an on-going basis which is generally the
assumed method for providing SLS compliance reporting and more broadly SLA and performance
reporting to both internal stakeholders such as engineering, network operations and service delivery as
well as external stakeholders, namely the subscriber.

Proactive on-going sessions deliver comprehensive monitoring at the expense of PM session traffic.
Therefore, proactive sessions can be supplemented with on-demand sessions to minimize the total
amount of PM session traffic. An on-demand PM session could be instantiated for specific testing
purposes, or where a performance attribute is not specified as a performance objective within the SLS
but the attribute is of interest to engineering or other stakeholders. For example, One-way Inter-Frame
Delay Variation may not be specified in the SLS but the CSP may wish to monitor this metric using an on-
demand PM session.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 12 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

2.1.8 Required instrumentation

As defined in MEF 35.1, the calculation of performance metrics is the responsibility of the network
element layer. To successfully support the implementation agreement, the network equipment must
support the various PM tools based on ITU-T Y.1731 and be able to perform the various calculations for
the associated performance metrics. Although not mandatory, the MEF has provided a SNMP-based MIB
in MEF specification 36.1 that reflects the requirements defined in the MEF 35.1 implementation
agreement. If one’s network equipment is not using this MIB, then the defined metrics must be accessed
using an EMS or an equipment-vendor-defined method. MEPs must be defined in the network elements.

Typically, that is done in one of the following types of network element:

Switch/router
Often a multi-service device with both layer 2 and layer 3 processing capabilities, these may be
used inside the operator’s network and in some cases at the customer edge. Due to their
greater functional capabilities, these elements are often costlier.

Network Interface Device (NID)

These are special purposed elements designed to terminate CE services and perform the end-to-
end performance metric calculations. Some provide both layer 2 and layer 3 termination
services. Given their specialized nature, they are often used to easily define all of the UNI
attributes of an EVC as well as setup the MEPs. They can be reasonably small and in fact recently
many vendors have introduced small-form pluggable (SFP) modules that can be inserted into
other network elements helping reduce the power and space requirements.

Test equipment
Various specialized test equipment including handheld and larger devices provide what is
typically the most comprehensive set of available performance metric calculations and generally
provide excellent support of the MEF defined best practices. However, as they are dedicated to
testing, they are not generally available for operational, that is proactive PM.

At minimum, the network equipment needs to support measurement intervals of 15 minutes as well as
on-demand durations of 1 minute to 15 minutes in 1 minute intervals. Furthermore, the network
equipment must be able to store 8 hours of historic (past) measurement data for each PM session for
any on-going (versus on-demand) PM sessions. This history must be stored per measurement interval,
that is, it cannot be aggregated. Ideally, the network equipment should be able to store up to 24 hours
of historic data.

2.1.8.1 Measurement Bins

To alleviate some storage and additional processing, the use of measurement bins is implemented
meaning the individual calculations made are not uniquely stored but are counted as part of a set of
intervals using measurement bins. Measurement bins are used to record the results of frame delay and
inter-frame delay variation measurements. These results can determine if the performance objectives
for FD (frame delay), FDR (frame delay range) and IFDV (inter-frame delay variation) are met.
Implementations (for all frame delay performance attributes) must support at least two bins and MEF
recommends at least ten.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 13 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
For example, bins may be created for frame delay with two intervals: (1) between 0ms and 100ms and
(2) 100 ms and higher, and a counter is used for each bin to count the number of times a FD
measurement was found within the respective interval. Separate bins and associated counters are used
for each measurement interval.

2.1.9 Threshold Crossing Alerts (TCAs)

The MEF also defines the ability to generate threshold crossing alerts northbound to the EMS or NMS
based on configured threshold levels for a particular PM session. This feature is optional. TCA reporting
can be stateless which means each and every time a threshold is crossed at least once during a single
measurement interval an alert is generated. Alternatively, TCAs can be stateful such that an alert is
received only for the first measurement interval the threshold is crossed and no further alerts will be
generated until a measurement interval occurs where the measurement is below the threshold, thus
clearing the alert. An alert is received for the 1st measurement interval in a sequence of intervals where
the configured threshold was crossed. Note that a change in availability state (from available to
unavailable or vice versa) is required functionality for the one-way availability performance attribute
and therefore there is no need to use TCA for availability.

2.1.10 Performance Objectives and CoS

Carrier Ethernet support multiple classes of service. The typical classes are H (high), M (medium) and L
(low). Performance objectives may be defined for each CoS defined in the EVC, as well as the bandwidth
profile making each CoS configurable to the specific needs of the network traffic it needs to carry.

Furthermore, given the technology agnostic approach of CE services, an operator can implement a CE
service over various technologies, access speeds and geographic distances. A performance objective that
is defined for a particular class of services is known as a CoS Performance objective or CPO. A
Performance Tier (PT) is a set of CPOs and may apply to the EVC as well as any OVCs that are part of the
EVC. MEF defines four PTs, based primarily on distance between UNIs with an associated one-way mean
frame delay.

 PT1 (Metro) <250km, MFD=2ms

 PT2 (Regional) <1200km, MFD=8ms
 PT3 (Continental) <7000km, MFD=44ms
 PT4 (Global) <27500km, MFD=172ms

As a result, each PT has a unique set of CPOs that is reflective of the distance across the network, in
particular with respect to delay measurements which are impacted by distance. An operator can
alternatively select the appropriate PT based on link speed (i.e. for low speed, copper-based
connectivity). Essentially the PTs can be considered a group of CPOs for a particular use case. That use
case may be based on distance, link speed, link technology or some other set of parameters. For each PT
defined, the MEF provides the respective CPOs which can be used based on the required PT and the
specific CoS (H – high, M – medium and L – Low). [please refer to MEF 23.2, Tables 9 through 12].

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 14 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

2.1.11 Practical Considerations and Terminology

2.1.11.1 Terminology Caveats

There is a lot of seemingly synonymous terminology within the realm of performance monitoring, and
the MEF specifications in particular. It is therefore helpful to recognize the sometimes subtle yet critical
differences in terms. The following table highlights some of these key differences:

Table 7 - Performance Terminology

Term Description Notes

Carrier Ethernet services have a number of
defined service attributes including
performance attributes. All of the service
Performance A specific characteristic of a Carrier attributes are specified in MEF 10.3 “Ethernet
Attribute3 Ethernet service’s performance. Service Attributes Phase 3”. Service attributes
define many aspects of the service such as the
type of service, the UNIs involved, bandwidth
profiles, etc.
This term is essentially synonymous with
performance attribute in the context of
defining the various performance attributes
Defined in MEF 10.3 as a such as one-way Frame Delay Performance. In
Performance
quantitative characterization of theory, the performance metric is defined
Metric4
Service Frame delivery quality. using actual service frames, that is, the real
traffic within the Carrier Ethernet service. In
practice, many of the performance metrics
are calculated using test traffic.
Without an objective defined, a service can
still be monitored and the performance
measured to provide information on the
The specific objective related to a
Performance performance attributes however, there is no
performance attribute that is
Objective agreement in place between subscriber and
defined in the SLS.
provider to deliver the service according to a
particular performance attribute(s) unless an
objective is defined within the SLS.

3
This paper uses the term ‘performance attribute’ when referring to the definition of the attribute as
opposed to a specific instance of the attribute.
4
This paper uses the term ‘performance metric’ when referring to a specific instance of a performance
attribute; i.e. the associated parameters are defined and a result is calculated.
MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 15 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

Term Description Notes

In most cases, the measurements used to
calculate a performance metric can only
approximate the true value of said
The measurement of a service’s
performance metric. The reason being that
performance. These
most measurements are based on test frames
measurements are used to
which are generally more practical than
Performance calculate the performance metric.
measuring actual service frames. In addition,
Measurement Many performance metrics require
such test measurements can be simpler to
multiple measurements to
implement and in some cases are the only
calculate the performance metric
available measurement option for a
of interest.
performance attribute (such as One-way
Frame Loss Ratio Performance in a
multipoint-to-multipoint network).

2.1.11.2 Definitions and settings

There are many decisions required by the operator configuring performance objectives and defining the
SLS. In particular, these include the time period applicable to the performance objective defined within
the SLS, e.g. availability must be a minimum of 99.99% over a month long period.

2.1.11.2.1 Defining the SLS

The CSP must define the SLS by identifying the following:
 Which (if any) performance attributes will be used (currently 10 including One-way Group
Availability and CPM). If none are used, then the remaining items are irrelevant.
 One (or more) performance objective(s) for each of the chosen performance attributes (recall
that different objectives can be defined for the same performance attribute as long as one or
more of their parameters differ, such as 80% of all service frames mush have FD ≤ 100ms and
90% of all service frames must have FD ≤ 200ms.)
 A set of UNI pairs within the EVC for which the objective is applied (these are uni-directional and
may be a subset of the total permutations of UNI pairs in the EVC)
 The related parameters for the objective(s) including but not limited to:
o The time period T to use for measuring adherence to each objective.
o The small time period Δt, for measuring FLR used to determine available time and
therefore whether any of the performance objectives are defined, as well as measuring
FLR and Availability if they are used in the SLS.
o The number of consecutive time intervals to use for calculating One-way Availability (n)
o The FLR that determines unavailability (C)
o The CoS ID(s) related to the objective
o The relevant percentile(s) for each objective

2.1.11.2.2 Configuring the performance measurements

Furthermore, the operator must define PDU sizes for the test frames, decide upon which PM solution(s)
to use to calculate the performance metric along with the measurement bins and related intervals for
bin-based counters. Finally, for all synthetic tests, the operator must determine the number of test PDUs

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 16 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
to be sent ensuring a statistically valid quantity, as well as how frequently they are sent. Both of which
have strong bearing on the ability to approximate the real service frame performance.

Many of these decisions may in fact be dictated by the available instrumentation (e.g. the capabilities
and adherence to standards within the network equipment itself). Many of the considerations and
caveats, as well as suggested default values for configuring the performance monitoring are included in
the MEF specifications themselves but are outside the scope of this paper.

2.1.11.3 Collection and reporting

After defining the objectives and configuring the network to deliver the necessary instrumentation, the
Carrier Ethernet provider must also ensure they have the ability to collect, aggregate and display the
resulting performance intelligence to the various stakeholders including network operations (in near real
time), engineering and planning (for both troubleshooting, design work, right-sizing, management
reporting), quality assurance and customer care and of course the end subscriber, whether a retail
business enterprise or a carrier or wireless operator in the way of SLA reporting and end-to-end
performance visibility.

A number of elements support the successful Carrier Ethernet provider in achieving these capabilities:

a. Instrumentation and management systems supporting the MEF standards (e.g. supporting MEF
35.1/36.1 where appropriate)
b. Capacity and scale to address real-time collection, query and display of service performance
metrics for large carrier-based networks
c. Role-based interface into the performance data to support the specific needs of each
stakeholder both inside the operator as well as external to the operator, e.g. subscriber-
oriented SLA reports,
d. Tailored reporting, preferably with APIs to enable on-demand report generation and analysis
e. Metric normalization to support legacy instrumentation that may be migrating to MEF-based
standards but not yet available in existing areas of the network
f. Service-based, multi-tenancy to map the collected metrics to the appropriate service (EVC or
OVC) and customer while maintaining secure access to data to authorized users
g. Direct, off-line access to performance data to enable subscribers to download their performance
metrics
h. Automated provisioning of collection and reporting in-line with Service Operations Functionality
within the framework of LSO

3. Summary
Assurance is one of the three tenets of the Third Network vision and truly distinguishes the concept of
the Third Network from that of the best-effort Internet. Thus service assurance is a driving force within
the work of the MEF. From creating a service OAM layer independent from the underlying network
transport technology, to defining the performance attributes necessary to monitor and manage a Carrier
Ethernet network, the fifty-plus specifications of the MEF often touch on this complex topic. From
service activation testing, performance management to fault management, there is a broad body of
work available to help the Carrier Ethernet provider implement Carrier Ethernet service assurance best
practices.
MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 17 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016

It begins with setting objectives and expectations between the provider of the service and the
subscriber of the service, covered in the companion paper, “Part I: An Introduction to Service Assurance
and Carrier Ethernet Service Performance Attributes” and then performing the necessary monitoring of
the network performance in order to determine whether the objectives are met. Furthermore, the
successful provider needs to deliver the service performance visibility to enhance the subscriber
experience and aid in both subscriber-based and provider-based decision making. From the MEF’s
definition of performance attributes, the service level specification and the PM solutions to the recently
published Carrier Ethernet Performance Reporting Framework specification (MEF 52), Carrier Ethernet
providers can best deliver a high-value, differentiated business class CE 2.0 service.

The MEF’s best practices leverage considerable work from the industry including the use of ITU-T’s
Recommendation G.8013/Y.1731 and the TM Forum’s SID model to name a few, and have developed a
rich set of attributes, guidelines and best practices to create resilient, valuable, business-class
connectivity services. The greater the level of adoption by all members of the industry (equipment
providers, LSO/OSS solution providers and Carrier Ethernet service providers), the faster the time to
revenue the industry can enjoy. Particularly in terms of launching Carrier Ethernet products with the
end-to-end SLAs the market is demanding. Furthermore, in the context of multi-operator services, as
each operator aligns on the same metrics, and begins to align on the display and sharing of the
performance data through standard APIs, the more effective and transparent multi-operator service
management becomes by agreeing on what to measure, how to measure and how to share those
measurements.

4. Guide to the MEF Technical Specifications

A summary of each specification that defines elements of performance management is provided for the
reader that wishes to investigate the technical details of each particular specification.

MEF 6.2 - EVC Ethernet Services Definitions Phase 3

Defines the various CE services and their attributes (E-Line, E-LAN, etc.) including the EVC performance
as well as envelopes and token sharing service attributes that relate to bandwidth profiles.

MEF 7.2 - Carrier Ethernet Information Model

This specification defines the EMS-NMS information model such that management system developers
can create interoperable CE service management systems.

MEF 10.3 - Ethernet Services Attributes Phase 3

Defines the service attributes for end-to-end Carrier Ethernet services, including the definition of
bandwidth profiles and the performance attributes used to define the SLS.

MEF 10.3.1 - Composite Performance Metric (CPM) Amendment to MEF 10.3

Defines an additional performance attribute; the composite performance metric (CPM).

MEF 15 - Requirements for Management of Metro Ethernet Phase 1 Network Elements

Original definition of statistics to be collected by network equipment.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 18 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
MEF 17 - Service OAM Framework and Requirements
As Carrier Ethernet services are technology agnostic, a common OAM definition is needed, regardless of
the transport layer, hence the Service OAM (SOAM) architectures was created as part of this
specification. Performance attributes defined within this specification have evolved in as part of more
recent specifications.

MEF 22.2 – Mobile Backhaul Phase 3 Implementation Agreement

CoS Mapping and CPOs and performance recommendations for mobile backhaul networks.

MEF 23.2 – Class of Service Phase 3 Implementation Agreement

Defines 3 specific classes of service (H, M and L), and introduces the concept of performance tiers (PT).
Specific values for performance objectives related to each PT and CoS are defined, as well as
recommended values for specific applications and their CoS. The framework presented can be used to
extend the MEF CoS model from the original 3 classes (H, M and L).

MEF 26.2 - External Network to Network Interface (ENNI) and Operator Service Attributes
This specification focuses on ENNI and defines the service attributes for inter-provider Carrier Ethernet
services. It includes the performance metrics and service level specification (SLS) for operator virtual
connections (OVC), as well as the definition of bandwidth profiles for inter-provider services.

MEF 33 - Ethernet Access Service Definition

This specification defines the Ethernet Access Services, that is, ENNI to UNI Carrier Ethernet services in
contrast to the EVC-based (UNI to UNI) services which are defined in MEF 6.2.

MEF 35.1 - Service OAM Performance Monitoring Implementation Agreement

Implementation agreement on performance monitoring; in essence the MEF’s best practices for
performance monitoring of CE services on which a significant portion of this white paper is based.

MEF 36.1 - Service OAM SNMP MIB for Performance Monitoring and MEF 39 - SOAM
Performance Monitoring YANG Module
MEF 36.1 is the instantiation of MEF 35.1 using the Simple Network Management Protocol (SNMP)’s MIB
structure. MEF 39 reflects the MEF 35 (predecessor to 35.1) using YANG models.

MEF 41 – Generic Token Bucket Algorithm

A specification of the Generic Token Bucket Algorithm based on the Bandwidth Profile Algorithm which
is specified in MEF 10.3.

MEF 51 - OVC Services Definition

This specification focuses on Operator Virtual Connection (OVC) Services based mainly on the service
attributes defined in MEF 26.1 (predecessor to 26.2); there are also some service attributes defined in
this document that go beyond MEF 26.1.

MEF 52 - Carrier Ethernet Performance Reporting Framework

Definition of the framework to support CE service subscribers with access to KPI performance data and
statistics through a standardized interface from their service provider. The specification borrows from
the TM Forum frameworks.

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 19 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.
Understanding Carrier Ethernet Service Assurance – Part II August 2016
MEF 55 - Lifecycle Service Orchestration (LSO): Reference Architecture and Framework
Details the LSO reference architecture and places the concept of service quality management, customer
performance reporting, performance events defined as part of service problem management.

5. About the MEF

MEF is the driving force enabling Third Network services for the digital economy and the hyper-connected
world. The Third Network concept combines the agility and ubiquity of the Internet with the assurance
and security of CE 2.0 (Carrier Ethernet 2.0). Third Network services provide an on-demand, orchestrated,
and assured experience with user-directed control over network resources and cloud connectivity. Third
Network services are delivered over automated, virtualized, and interconnected networks powered by
LSO, SDN, NFV, and CE 2.0.

MEF leverages its global 210+ network operators and technology vendor community, builds upon the
robust $80 billion Carrier Ethernet market, and provides a practical evolution to the Third Network with
LSO, SDN, and NFV implementations that build upon a CE 2.0 foundation. For more information, see
www.MEF.net.

6. Glossary and Terms

A glossary of terms used in this document can be found online at MEF.net.

7. References
Source Link
MEF The Third Network, Vision and Strategy (Based on Network as a Service Principles)
MEF CE 2.0 Service Management Life Cycle White Paper
MEF Published Technical Specifications
Understanding Carrier Ethernet Service Assurance - Part I: An Introduction to
MEF
Service Assurance and Carrier Ethernet Service Performance Attributes
Recommendation G.8013/Y.1731: Operation, administration and maintenance
ITU-T
(OAM) functions and mechanisms for Ethernet-based networks

8. Acknowledgements
Editor: Christopher Cullan, InfoVista
Principal Authors:
 Christopher Cullan, InfoVista
Contributors:
 David Ball, Cisco
 Yoav Cohen, RAD
 Eric Leser, Charter Communications
 John Siwko, AT&T
 Marjory Sy, PLDT

MEF © MEF Forum 2016. Any reproduction of this document, or any portion thereof, shall contain the Page 20 of 20
following statement: "Reproduced with permission of the MEF Forum." No user of this document is
2016026 authorized to modify any of the information contained herein.