Neighbor discovery enables each end of a link to learn the OAM capabilities of the other end and establish an OAM peer relationship. Each end also can require that the peer have certain capabilities before it will establish a session. You can configure certain actions to be taken if there is a capabilities conflict or if a discovery process times out, using the action capabilities-conflict or action discovery-timeout commands.

Link monitoring enables an OAM peer to monitor faults that cause the quality of a link to deteriorate over time. When link monitoring is enabled, an OAM peer can be configured to take action when the configured thresholds are exceeded.

MIB retrieval enables an OAM peer on one side of an interface to get the MIB variables from the remote side of the link. The MIB variables that are retrieved from the remote OAM peer are READ ONLY.

Miswiring Detection is a Cisco-proprietary feature that uses the 32-bit vendor field in every Information OAMPDU to identify potential miswiring cases.

SNMP traps can be enabled or disabled on an Ethernet OAM interface.

A maintenance domain describes a management space for the purpose of managing and administering a network. A domain is owned and operated by a single entity and defined by the set of interfaces internal to it and at its boundary, as shown in this figure.

A maintenance domain is defined by the bridge ports that are provisioned within it. Domains are assigned maintenance levels, in the range of 0 to 7, by the administrator. The level of the domain is useful in defining the hierarchical relationships of multiple domains.

CFM maintenance domains allow different organizations to use CFM in the same network, but independently. For example, consider a service provider who offers a service to a customer, and to provide that service, they use two other operators in segments of the network. In this environment, CFM can be used in the following ways:

• The customer can use CFM between their CE devices, to verify and manage connectivity across the whole network.

• The service provider can use CFM between their PE devices, to verify and manage the services they are providing.

• Each operator can use CFM within their operator network, to verify and manage connectivity within their network.

Each organization uses a different CFM maintenance domain.

This figure shows an example of the different levels of maintenance domains in a network.

Note	In CFM diagrams, the conventions are that triangles represent MEPs, pointing in the direction that the MEP sends CFM frames, and circles represent MIPs. For more information about MEPs and MIPs, see the “Maintenance Points” section on page 71.

To ensure that the CFM frames for each domain do not interfere with each other, each domain is assigned a maintenance level, between 0 and 7. Where domains are nested, as in this example, the encompassing domain must have a higher level than the domain it encloses. In this case, the domain levels must be negotiated between the organizations involved. The maintenance level is carried in all CFM frames that relate to that domain.

CFM maintenance domains may touch or nest, but cannot intersect. This figure illustrates the supported structure for touching and nested domains, and the unsupported intersection of domains.

A CFM service allows an organization to partition its CFM maintenance domain, according to the connectivity within the network. For example, if the network is divided into a number of virtual LANs (VLANs), a CFM service is created for each of these. CFM can then operate independently in each service. It is important that the CFM services match the network topology, so that CFM frames relating to one service cannot be received in a different service. For example, a service provider may use a separate CFM service for each of their customers, to verify and manage connectivity between that customer's end points.

A CFM service is always associated with the maintenance domain that it operates within, and therefore with that domain's maintenance level. All CFM frames relating to the service carry the maintenance level of the corresponding domain.

Note	CFM Services are referred to as Maintenance Associations in IEEE 802.1ag and as Maintenance Entity Groups in ITU-T Y.1731.

A CFM Maintenance Point (MP) is an instance of a particular CFM service on a specific interface. CFM only operates on an interface if there is a CFM maintenance point on the interface; otherwise, CFM frames are forwarded transparently through the interface.

A maintenance point is always associated with a particular CFM service, and therefore with a particular maintenance domain at a particular level. Maintenance points generally only process CFM frames at the same level as their associated maintenance domain. Frames at a higher maintenance level are always forwarded transparently, while frames at a lower maintenance level are normally dropped. This helps enforce the maintenance domain hierarchy described in the “Maintenance Domains” section on page 69, and ensures that CFM frames for a particular domain cannot leak out beyond the boundary of the domain.

There are two types of MP:

• Maintenance End Points (MEPs)—Created at the edge of the domain. Maintenance end points (MEPs) are members of a particular service within a domain and are responsible for sourcing and sinking CFM frames. They periodically transmit continuity check messages and receive similar messages from other MEPs within their domain. They also transmit traceroute and loopback messages at the request of the administrator. MEPs are responsible for confining CFM messages within the domain.

• Maintenance Intermediate Points (MIPs)—Created in the middle of the domain. Unlike MEPS, MIPs do allow CFM frames at their own level to be forwarded.

Unlike MEPs, MIPs are not explicitly configured on each interface. MIPs are created automatically according to the algorithm specified in the CFM 802.1ag standard. The algorithm, in brief, operates as follows for each interface:

• The cross-connect for the interface is found, and all services associated with that cross-connect are considered for MIP auto-creation.

• The level of the highest-level MEP on the interface is found. From among the services considered above, the service in the domain with the lowest level that is higher than the highest MEP level is selected. If there are no MEPs on the interface, the service in the domain with the lowest level is selected.

• The MIP auto-creation configuration (mip auto-create command) for the selected service is examined to determine whether a MIP should be created.

  Note	Configuring a MIP auto-creation policy for a service does not guarantee that a MIP will automatically be created for that service. The policy is only considered if that service is selected by the algorithm first.

The boundary of a domain is an interface, rather than a bridge or host. Therefore, MEPs can be sub-divided into two categories:

• Down MEPs—Send CFM frames from the interface where they are configured, and process CFM frames received on that interface. Down MEPs transmit AIS messages upward (toward the cross-connect).

• Up MEPs—Send frames into the bridge relay function, as if they had been received on the interface where the MEP is configured. They process CFM frames that have been received on other interfaces, and have been switched through the bridge relay function as if they are going to be sent out of the interface where the MEP is configured. Up MEPs transmit AIS messages downward (toward the wire). However, AIS packets are only sent when there is a MIP configured on the same interface as the MEP and at the level of the MIP.

Note	The terms Down MEP and Up MEP are defined in the IEEE 802.1ag and ITU-T Y.1731 standards, and refer to the direction that CFM frames are sent from the MEP. The terms should not be confused with the operational status of the MEP.

This figure illustrates the monitored areas for Down and Up MEPs.

This figure shows maintenance points at different levels. Because domains are allowed to nest but not intersect (see Figure 3), a MEP at a low level always corresponds with a MEP or MIP at a higher level. In addition, only a single MIP is allowed on any interface—this is generally created in the lowest domain that exists at the interface and that does not have a MEP.

MIPs and Up MEPs can only exist on switched (Layer 2) interfaces, because they send and receive frames from the bridge relay function. Down MEPs can be created on switched (Layer 2) or routed (Layer 3) interfaces.

MEPs continue to operate normally if the interface they are created on is blocked by the Spanning Tree Protocol (STP); that is, CFM frames at the level of the MEP continue to be sent and received, according to the direction of the MEP. MEPs never allow CFM frames at the level of the MEP to be forwarded, so the STP block is maintained.

MIPs also continue to receive CFM frames at their level if the interface is STP blocked, and can respond to any received frames. However, MIPs do not allow CFM frames at the level of the MIP to be forwarded if the interface is blocked.

Note

A separate set of CFM maintenance levels is created every time a VLAN tag is pushed onto the frame. Therefore, if CFM frames are received on an interface which pushes an additional tag, so as to “tunnel” the frames over part of the network, the CFM frames will not be processed by any MPs within the tunnel, even if they are at the same level. For example, if a CFM MP is created on an interface with an encapsulation that matches a single VLAN tag, any CFM frames that are received at the interface that have two VLAN tags will be forwarded transparently, regardless of the CFM level.

The CFM protocol consists of a number of different message types, with different purposes. All CFM messages use the CFM EtherType, and carry the CFM maintenance level for the domain to which they apply.

This section describes the following CFM messages:

Continuity Check Messages (CCMs) are “heartbeat” messages exchanged periodically between all the MEPs in a service. Each MEP sends out multicast CCMs, and receives CCMs from all the other MEPs in the service—these are referred to as peer MEPs. This allows each MEP to discover its peer MEPs, and to verify that there is connectivity between them.

MIPs also receive CCMs. MIPs use the information to build a MAC learning database that is used when responding to Linktrace. For more information about Linktrace, see the “Linktrace (IEEE 802.1ag and ITU-T Y.1731)” section on page 77.

All the MEPs in a service must transmit CCMs at the same interval. IEEE 802.1ag defines 7 possible intervals that can be used:

• 10ms (applicable on the Cisco ASR 9000 Enhanced Ethernet Line Card)

• 100ms

• 1s

• 10s

• 1 minute

• 10 minutes

A MEP detects a loss of connectivity with one of its peer MEPs when some number of CCMs have been missed. This occurs when sufficient time has passed during which a certain number of CCMs were expected, given the CCM interval. This number is called the loss threshold, and is usually set to 3.

CCM messages carry a variety of information that allows different defects to be detected in the service. This information includes:

• A configured identifier for the domain of the transmitting MEP. This is referred to as the Maintenance Domain Identifier (MDID).

• A configured identifier for the service of the transmitting MEP. This is referred to as the Short MA Name (SMAN). Together, the MDID and the SMAN make up the Maintenance Association Identifier (MAID). The MAID must be configured identically on every MEP in the service.

• A configured numeric identifier for the MEP (the MEP ID). Each MEP in the service must be configured with a different MEP ID.

• A sequence number.

• A Remote Defect Indication (RDI). Each MEP includes this in the CCMs it is sending, if it has detected a defect relating to the CCMs it is receiving. This notifies all the MEPs in the service that a defect has been detected somewhere in the service.

• The interval at which CCMs are being transmitted.

• The status of the interface where the MEP is operating—for example, whether the interface is up, down, STP blocked, and so on.

  Note	The status of the interface (up/down) should not be confused with the direction of any MEPs on the interface (Up MEPs/Down MEPs).

These defects can be detected from received CCMs:

• Interval mismatch—The CCM interval in the received CCM does not match the interval that the MEP is sending CCMs.

• Level mismatch—A MEP has received a CCM carrying a lower maintenance level than the MEPs own level.

• Loop—A CCM is received with the source MAC address equal to the MAC address of the interface where the MEP is operating.

• Configuration error—A CCM is received with the same MEP ID as the MEP ID configured for the receiving MEP.

• Cross-connect—A CCM is received with an MAID that does not match the locally configured MAID. This generally indicates a VLAN misconfiguration within the network, such that CCMs from one service are leaking into a different service.

• Peer interface down—A CCM is received that indicates the interface on the peer is down.

• Remote defect indication—A CCM is received carrying a remote defect indication.

  Note	This defect does not cause the MEP to include a remote defect indication in the CCMs that it is sending.

Out-of-sequence CCMs can also be detected by monitoring the sequence number in the received CCMs from each peer MEP. However, this is not considered a CCM defect.

Loopback Messages (LBM) and Loopback Replies (LBR) are used to verify connectivity between a local MEP and a particular remote MP. At the request of the administrator, a local MEP sends unicast LBMs to the remote MP. On receiving each LBM, the target maintenance point sends an LBR back to the originating MEP. Loopback indicates whether the destination is reachable or not—it does not allow hop-by-hop discovery of the path. It is similar in concept to an ICMP Echo (ping). Since loopback messages are destined for unicast addresses, they are forwarded like normal data traffic, while observing the maintenance levels. At each device that the loopback reaches, if the outgoing interface is known (in the bridge's forwarding database), then the frame is sent out on that interface. If the outgoing interface is not known, then the message is flooded on all interfaces.

This figure shows an example of CFM loopback message flow between a MEP and MIP.

Loopback messages can be padded with user-specified data. This allows data corruption to be detected in the network. They also carry a sequence number which allows for out-of-order frames to be detected.

Except for one-way delay and jitter measurements, loopback messages can also be used for Ethernet SLA, if the peer does not support delay measurement.

Linktrace Messages (LTM) and Linktrace Replies (LTR) are used to track the path (hop-by-hop) to a unicast destination MAC address. At the request of the operator, a local MEP sends an LTM. Each hop where there is a maintenance point sends an LTR back to the originating MEP. This allows the administrator to discover connectivity data about the path. It is similar in concept to IP traceroute, although the mechanism is different. In IP traceroute, successive probes are sent, whereas CFM Linktrace uses a single LTM which is forwarded by each MP in the path. LTMs are multicast, and carry the unicast target MAC address as data within the frame. They are intercepted at each hop where there is a maintenance point, and either retransmitted or dropped to discover the unicast path to the target MAC address.

This figure shows an example of CFM linktrace message flow between MEPs and MIPs.

The linktrace mechanism is designed to provide useful information even after a network failure. This allows it to be used to locate failures, for example after a loss of continuity is detected. To achieve this, each MP maintains a CCM Learning Database. This maps the source MAC address for each received CCM to the interface through which the CCM was received. It is similar to a typical bridge MAC learning database, except that it is based only on CCMs and it times out much more slowly—on the order of days rather than minutes.

Note	In IEEE 802.1ag, the CCM Learning Database is referred to as the MIP CCM Database. However, it applies to both MIPs and MEPs.

In IEEE 802.1ag, when an MP receives an LTM message, it determines whether to send a reply using the following steps:

1. The target MAC address in the LTM is looked up in the bridge MAC learning table. If the MAC address is known, and therefore the egress interface is known, then an LTR is sent.

2. If the MAC address is not found in the bridge MAC learning table, then it is looked up in the CCM learning database. If it is found, then an LTR is sent.

3. If the MAC address is not found, then no LTR is sent (and the LTM is not forwarded).

If the target MAC has never been seen previously in the network, the linktrace operation will not produce any results.

Note

IEEE 802.1ag and ITU-T Y.1731 define slightly different linktrace mechanisms. In particular, the use of the CCM learning database and the algorithm described above for responding to LTM messages are specific to IEEE 802.1ag. IEEE 802.1ag also specifies additional information that can be included in LTRs. Regardless of the differences, the two mechanisms are interoperable.

Exploratory Linktrace is a Cisco extension to the standard linktrace mechanism described above. It has two primary purposes:

• Provide a mechanism to locate faults in cases where standard linktrace does not work, such as when a MAC address has never been seen previously in the network. For example, if a new MEP has been provisioned but is not working, standard linktrace does not help isolate a problem because no frames will ever have been received from the new MEP. Exploratory Linktrace overcomes this problem.

• Provide a mechanism to map the complete active network topology from a single node. This can only be done currently by examining the topology (for example, the STP blocking state) on each node in the network individually, and manually combining this information to create the overall active topology map. Exploratory linktrace allows this to be done automatically from a single node.

Exploratory Linktrace is implemented using the Vendor Specific Message (VSM) and Vendor Specific Reply (VSR) frames defined in ITU-T Y.1731. These allow vendor-specific extensions to be implemented without degrading interoperability. Exploratory Linktrace can safely be deployed in a network that includes other CFM implementations because those implementations will simply ignore the Exploratory Linktrace messages.

Exploratory Linktrace is initiated at the request of the administrator, and results in the local MEP sending a multicast Exploratory Linktrace message. Each MP in the network that receives the message sends an Exploratory Linktrace reply. MIPs that receive the message also forward it on. The initiating MEP uses all the replies to create a tree of the overall network topology.

This figure show an example of the Exploratory Linktrace message flow between MEPs.

To avoid overloading the originating MEP with replies in a large network, responding MPs delay sending their replies for a random amount of time, and that time increases as the size of the network increases.

In a large network, there will be a corresponding large number of replies and the resulting topology map will be equally large. If only a part of the network is of interest, for example, because a problem has already been narrowed down to a small area, then the Exploratory Linktrace can be “directed” to start at a particular MP. Replies will thus only be received from MPs beyond that point in the network. The replies are still sent back to the originating MEP.

The router supports one-way and two-way delay measurement using two packet types:

• Delay Measurement Message (DMM)

• Delay Measurement Response (DMR)

These packets are unicast similar to loopback messages. The packets carry timestamps generated by the system time-of-day clock to support more accurate delay measurement, and also support an SLA manageability front-end.

However, unlike loopback messages, these message types can also measure one-way delay and jitter either from destination to source, or from source to destination.

For more information about SLA, see the “Ethernet SLA” section on page 87.

Synthetic Loss Measurement (SLM) is a mechanism that injects synthetic measurement probes, and measures the loss of these probes in order to measure the loss of real data traffic. Each probe packet carries a sequence number, and the sender increments the sequence number by one for each packet that is sent and the receiver can thereby detect the lost packets by looking for missing sequence numbers.

SLM packets contain two sequence numbers; one written by the initiator into the SLM and copied by the responder into the SLR, and the other allocated by the responder and written into the SLR. These are refered to as the source-to-destination (sd) sequence number and the destination-to-source (ds) sequence number respectively.

This figure shows an example of how the sequence numbers are used to calculate the Frame Loss Ratio (FLR) in each direction.

Y.1731 Loss Measurement is a mechanism that measures the actual data traffic loss between a pair of MEPs in a point-to-point Ethernet service. This is in contrast to the Synthetic Loss Measurement, which measures the frame loss of synthetic frames. By using Y.1731 Loss Measurement, you can measure the one-way loss in each direction, for each priority class and also measure the loss aggregated across all priority classes.

To enable loss measurements to be made, each MEP maintains, for each priority class, both source-to-destination and destination-to-source frame counts for its peer MEPs.

There are two Loss Measurement Mechanisms (LMM); namely, single-ended and dual-ended. 
Cisco IOS XR Software supports only single-ended LMM.

MEP cross-check supports configuration of a set of expected peer MEPs so that errors can be detected when any of the known MEPs are missing, or if any additional peer MEPs are detected that are not in the expected group.

The set of expected MEP IDs in the service is user-defined. Optionally, the corresponding MAC addresses can also be specified. CFM monitors the set of peer MEPs from which CCMs are being received. If no CCMs are ever received from one of the specified expected peer MEPs, or if a loss of continuity is detected, then a cross-check “missing” defect is detected. Similarly, if CCMs are received from a matching MEP ID but with the wrong source MAC address, a cross-check “missing” defect is detected. If CCMs are subsequently received that match the expected MEP ID, and if specified, the expected MAC address, then the defect is cleared.

Note	While loss of continuity can be detected for any peer MEP, it is only treated as a defect condition if cross-check is configured.

If cross-check is configured and CCMs are received from a peer MEP with a MEP ID that is not expected, this is detected as a cross-check “unexpected” condition. However, this is not treated as a defect condition.

CFM supports logging of various conditions to syslog. Logging can be enabled independently for each service, and when the following conditions occur:

• New peer MEPs are detected, or loss of continuity with a peer MEP occurs.

• Changes to the CCM defect conditions are detected.

• Cross-check “missing” or “unexpected” conditions are detected.

• AIS condition detected (AIS messages received) or cleared (AIS messages no longer received).

• EFD used to shut down an interface, or bring it back up.

Ethernet Fault Detection (EFD) is a mechanism that allows Ethernet OAM protocols, such as CFM, to control the “line protocol” state of an interface.

Unlike many other interface types, Ethernet interfaces do not have a line protocol, whose state is independent from that of the interface. For Ethernet interfaces, this role is handled by the physical-layer Ethernet protocol itself, and therefore if the interface is physically up, then it is available and traffic can flow.

EFD changes this to allow CFM to act as the line protocol for Ethernet interfaces. This allows CFM to control the interface state so that if a CFM defect (such as AIS or loss of continuity) is detected with an expected peer MEP, the interface can be shut down. This not only stops any traffic flowing, but also triggers actions in any higher-level protocols to route around the problem. For example, in the case of Layer 2 interfaces, the MAC table would be cleared and MSTP would reconverge. For Layer 3 interfaces, the ARP cache would be cleared and potentially the IGP would reconverge.

Note	EFD can only be used for down MEPs. When EFD is used to shut down the interface, the CFM frames continue to flow. This allows CFM to detect when the problem has been resolved, and thus bring the interface backup automatically.

This figure shows CFM detection of an error on one of its sessions EFD signaling an error to the corresponding MAC layer for the interface. This triggers the MAC to go to a down state, which further triggers all higher level protocols (Layer 2 pseudowires, IP protocols, and so on) to go down and also trigger a reconvergence where possible. As soon as CFM detects there is no longer any error, it can signal to EFD and all protocols will once again go active.

The ITU-T Y.1731 standard defines several mechanisms that can be used for performance monitoring in Carrier Ethernet networks. These are the measurement mechanisms that were defined in the standard:

Delay Measurement: This can be used to accurately measure frame delay by exchanging CFM frames containing timestamps, and to measure inter-frame delay variation (jitter) by comparing consecutive delay measurements. Delay Measurement messages can be used to perform these measurements:

• Round-trip time

• Round-trip Jitter

• One-way delay (both SD and DS)

• One-way jitter (both SD and DS)

• SLA Probe Packet corruption count

• Out of order SLA probe packet count

• SLA probe packet loss

Synthetic Loss Measurement: The loss measurement mechanism defined in Y.1731 can only be used in point-to-point networks, and only works when there is sufficient data traffic flowing. The difficulties with the Y.1731 Loss Measurement mechanism was recognized across the industry and hence an alternative mechanism has been defined and standardized for measuring loss.

This alternative mechanism does not measure the loss of the actual data traffic, but instead injects synthetic CFM frames and measures the loss of these synthetic frames. Statistical analysis can then be used to give an approximation to the loss of data traffic. This technique is called Synthetic Loss Measurement. This has been included in the latest version of the Y.1731 standard. Synthetic Loss Measurement messages can be used to perform these measurements:

• One-way loss (Source to Destination)

• One-way loss (Destination to Source)

Loopback: This is not primarily targetted at performance monitoring, but can be used to approximate round-trip delay and jitter, such as when the peer device does not support delay measurement. Loopback messages can be used to perform these measurements:

• Round-trip time

• Round-trip jitter

• SLA probe packet corruption count

• Out of order SLA probe packet count

• SLA probe packet loss

These are the commonly used terminology in Loss Measurement Mechanism:

• Single-ended: A mechanism where device A sends a measurement packet to device B, which in turn sends a response back to device A. All calculations and results are done on device A.

• Dual-ended: A mechanism where device A sends a measurement packet to device B, which does not send a response. All calculations and results are done on device B.

• One-way: A measurement of the performance of packets flowing in one direction, from device A to device B, or from device B to device A.

• Two-way: A measurement of the performance of packets flowing from device A to device B, and back to device A.

• Forwards: A one-way measurement from the initiator (device A) to the receiver, or responder (device B).

• Backwards: A one-way measurement from the responder (device B) to the initiator (device A).

Note	Cisco IOS XR Software supports only single-ended LMM.

These are two primary attributes that can be calculated based on loss measurements:

• Frame Loss Ratio (FLR)

• Availability

Frame Loss Ratio is the ratio of lost packets to sent packets:

(<num_sent > - <num_rcvd> )/(<num_sent>)

It is normally expressed as a percentage. The accuracy of the measurement depends majorly on the number of packets sent.

Availability is a complex attribute, typically measured over a long period of time, such as weeks or months. The intent of this performance attribute is to measure the proportion of time when there was prolonged high loss. Cisco IOS XR Software does not track the availability.

1. Data loss measurement cannot be used in a multipoint service; it can only be used in a peer-to-peer service.

2. As a Loss Measurement Reply (LMR) contains no sequence IDs, the only field, which can be used to distinguish to which probe a given LMR corresponds, is the priority level. Also, the priority level is the only field that can determine whether the LMR is in response to an on-demand or proactive operation. This limits the number of Loss Measurement probes that can be active at a time for each local MEP to 16.

3. As loss measurements are made on a per-priority class basis, QoS policies, which alter the priority of packets processed by the network element, or re-order packets can affect the accuracy of the calculations. For the highest accuracy, packets must be counted after any QoS policies have been applied.

4. The accuracy of data loss measurement is highly dependent on the number of data packets that are sent. If the volume of data traffic is low, errors with the packet counts might be magnified. If there is no data traffic flowing, no loss measurement performance attributes can be calculated. If aggregate measurements are taken, then only 2 probes can be active at the same time: one proactive and one on-demand.

5. The accuracy of data loss measurement is highly dependent on the accuracy of platform-specific packet counters. Due to hardware limitations, it may not be possible to achieve completely accurate packet counters, especially if QoS policies are applied to the packets being counted.

6. Performing data loss measurement can have an impact on the forwarding performance of network elements; this is because of the need to count, as well as forward the packets.

7. Before starting any LMM probes,it is necessary to allocate packet counters for use with LMM on both ends (assuming both ends are running Cisco IOS XR Software).

To successfully configure the Cisco Ethernet SLA feature, you should understand the following concepts:

A statistic in Ethernet SLA is a single performance parameter. These statistics can be measured by Ethernet SLA:

• Round-trip delay

• Round-trip jitter

• One-way delay from source to destination

• One-way jitter from source to destination

• One-way frame loss from source to destination

• One-way delay from destination to source

• One-way jitter from destination to source

• One-way frame loss from destination to source

Note	Not all statistics can be measured by all types of packet. For example, one-way statistics cannot be measured when using CFM loopback packets.

An Ethernet SLA measurement packet is a single protocol message and corresponding reply that is sent on the network for the purpose of making SLA measurements. These types of measurement packet are supported:

• CFM Delay Measurement (Y.1731 DMM/DMR packets)—CFM delay measurement packets contain timestamps within the packet data that can be used for accurate measurement of frame delay and jitter. These packets can be used to measure round-trip or one-way statistics; however, the size of the DMM/DMR packets cannot be modified.

Note

From Cisco IOS XR Release 4.3.x onwards, you can configure the Ethernet SLA profile to use Y.1731 DMM v1 frames. The restriction of 150 configured Ethernet SLA operations for each CFM MEP is removed not only for profiles using DMM frames, but also for profiles using the other supported Y.1731 frame types, such as loopback measurement and synthetic loss measurement. For interoperability purposes, it is still possible to configure operations to use DMM v0 frames. This is done by specifying a type of cfm-delay-measurement-v0 on the ethernet SLA profile command. The limit of 150 configured operations for each CFM MEP still applies in this case.

• CFM loopback (LBM/LBR)—CFM loopback packets are less accurate, but can be used if the peer device does not support DMM/DMR packets. Only round-trip statistics can be measured because these packets do not contain timestamps. However, loopback packets can be padded, so measurements can be made using frames of a specific size.

• CFM Synthetic Loss Measurement (Y.1731 SLM/SLR packets)—SLM packets contain two sequence numbers; one written by the initiator into the SLM and copied by the responder into the SLR, and the other allocated by the responder and written into the SLR. These are refered to as the source-to-destination (sd) sequence number and the destination-to-source (ds) sequence number respectively.

Note	Because SLM is a statistical sampling technique, there may be some variance of the measured value around the actual loss value. Also, the accuracy of the measurement is improved by using more SLM packets for each FLR calculation.

• CFM Loss Measurement (Y.1731 LMM/LMR packets)— As LMMs and LMRs contain no sequence ID, there is a limited set of data that can be used to distinguish different Loss Measurement operations, limiting the number of concurrent operations for each MEP.

A sample is a single result—a number—that relates to a given statistic. For some statistics such as round-trip delay, a sample can be measured using a single measurement packet. For other statistics such as jitter, obtaining a sample requires two measurement packets.

A probe is a sequence of measurement packets used to gather SLA samples for a specific set of statistics. The measurement packets in a probe are of a specific type (for example, CFM delay measurement or CFM loopback) and have specific attributes, such as the frame size and priority.

Note	A single probe can collect data for different statistics at the same time, using the same measurement packets (for example, one-way delay and round-trip jitter).

Within a probe, measurement packets can either be sent individually, or in bursts. A burst contains two or more packets sent within a short interval apart. Each burst can last up to one minute, and bursts can follow each other immediately to provide continuous measurement within the probe.

For statistics that require two measurement packets for each sample (such as jitter), samples are only calculated based on measurement packets in the same burst. For all statistics, it is more efficient to use bursts than to send individual packets.

Note	If bursts are configured back to back, so as to cause a continuous and uninterrupted flow of SLA packets, then packets at the end of one burst and the start of the next are used in Loss Measurement calculations.

An Ethernet SLA schedule describes how often probes are sent, how long each probe lasts, and at what time the first probe starts.

Note	If probes are scheduled back to back, so as to cause a continuous and uninterrupted flow of SLA packets, then packets at the end of one probe and the start of the next are used in Loss Measurement calculations.

For a particular statistic, a bucket is a collection of results that were gathered during a particular period of time. All of the samples for measurements that were initiated during the period of time represented by a bucket are stored in that bucket. Buckets allow results from different periods of time to be compared (for example, peak traffic to off-peak traffic).

By default, a separate bucket is created for each probe; that is, the bucket represents the period of time starting at the same time as the probe started, and continuing for the duration of the probe. The bucket will therefore contain all the results relating to measurements made by that probe.

Rather than storing each sample separately within a bucket, an alternative is to aggregate the samples into bins. An aggregation bin is a range of sample values, and contains a counter of the number of samples that were received that fall within that range. The set of bins forms a histogram. When aggregation is enabled, each bucket contains a separate set of bins. See this figure.

An operation profile is a configuration entity that defines the following aspects of an operation:

• What packet types to send and in what quantities (probe and burst configuration)

• What statistics to measure, and how to aggregate them

• When to schedule the probes

An operation profile by itself does not cause any packets to be sent or statistics collected, but is used to create operation instances.

An operation is an instance of a given operation profile that is actively collecting performance data. Operation instances are created by associating an operation profile with a given source (an interface and MEP) and with a given destination (a MEP ID or MAC address). Operation instances exist for as long as the configuration is applied, and they run for an indefinite duration on an ongoing basis.

An on-demand operation is a method of Ethernet SLA operation that can be run on an as-needed basis for a specific and finite period of time. This can be useful in situations such as when you are starting a new service or modifying the parameters for a service to verify the impact of the changes, or if you want to run a more detailed probe when a problem is detected by an ongoing scheduled operation.

On-demand operations do not use profiles and have a finite duration. The statistics that are collected are discarded after a finite time after the operation completes (two weeks), or when you manually clear them.

On-demand operations are not persistent so they are lost during certain events such as a card reload or Minimal Disruptive Restart (MDR).

Ethernet SLA statistics measurement for network performance is performed by sending packets and storing data metrics such as:

• Round-trip delay time—The time for a packet to travel from source to destination and back to source again.

• Round-trip jitter—The variance in round-trip delay time (latency).

• One-way delay and jitter—The router also supports measurement of one-way delay or jitter from source to destination, or from destination to source.

• One-way frame loss—The router also supports measurement of one-way frame loss from source to destination, or from destination to source.

In addition to these metrics, these statistics are also kept for SLA probe packets:

• Packet loss count

• Packet corruption event

• Out-of-order event

• Frame Loss Ratio (FLR)

Counters for packet loss, corruption and out-of-order packets are kept for each bucket, and in each case, a percentage of the total number of samples for that bucket is reported (for example, 4% packet corruption). For delay, jitter, and loss statistics, the minimum, maximum, mean and standard deviation for the whole bucket are reported, as well as the individual samples or aggregated bins. Also, the overall FLR for the bucket, and individual FLR measurements or aggregated bins are reported for synthetic loss measurement statistics. The packet loss count is the overall number of measurement packets lost in either direction and the one-way FLR measures the loss in each direction separately.

When aggregation is enabled using the aggregate command, bins are created to store a count of the samples that fall within a certain value range, which is set by the width keyword. Only a counter of the number of results that fall within the range for each bin is stored. This uses less memory than storing individual results. When aggregation is not used, each sample is stored separately, which can provide a more accurate statistics analysis for the operation, but it is highly memory-intensive due to the independent storage of each sample.

A bucket represents a time period during which statistics are collected. All the results received during that time period are recorded in the corresponding bucket. If aggregation is enabled, each bucket has its own set of bins and counters, and only results relating to the measurements initiated during the time period represented by the bucket are included in those counters.

By default, there is a separate bucket for each probe. The time period is determined by how long the probe lasts (configured by the probe, send (SLA), and schedule (SLA) commands).You can modify the size of buckets so that you can have more buckets per probe or fewer buckets per probe (less buckets allows the results from multiple probes to be included in the same bucket). Changing the size of the buckets for a given metric clears all stored data for that metric. All existing buckets are deleted and new buckets are created.

Scheduled SLA operation profiles run indefinitely, according to a configured schedule, and the statistics that are collected are stored in a rolling buffer, where data in the oldest bucket is discarded when a new bucket needs to be recorded.

Frame Loss Ratio (FLR) is a primary attribute that can be calculated based on loss measurements. FLR is defined by the ratio of lost packets to sent packets and expressed as a percentage value. FLR is measured in each direction (source to destination and destination to source) separately. Availability is an attribute, that is typically measured over a long period of time, such as weeks or months. The intent is to measure the proportion of time when there was prolonged high loss.

When you configure a scheduled Ethernet SLA operation, you perform these basic steps:

1. Configure global profiles to define how packets are sent in each probe, how the probes are scheduled, and how the results are stored.

2. Configure operations from a specific local MEP to a specific peer MEP using these profiles.

Note	Certain Ethernet SLA configurations use large amounts of memory which can affect the performance of other features on the system. For more information, see the “Configuring Ethernet SLA” section on page 128.

The Ethernet Data Plane Loopback (EDPL) is implemented on the Satellite nV System as shown in this illustration.

The internal and external EDPL are realized as follows:

• Internal Loopback: The MAC address swap happens on the host and the frame actually gets looped back from the satellite where Layer 1 loopback needs to be turned on at the port. As the entire port is looped back on the satellite, the internal loopback for satellite ports cannot loopback or filter specific sub-interface sessions on the port. You need to enable both EDPL and port L1 internal loopback on the satellite port for this functionality.

• External Loopback: The external loopback is currently implemented entirely on the host because of the need to perform MAC address swap.

The support that the Ethernet Data Plane Loopback feature provides is:

• Locally-enabled Ethernet Data Plane Loopback on all Ethernet interface types, such as physical and bundle interfaces and sub-interfaces.

• In the case of Layer 2 interfaces, support for these types of looping back of traffic:

  • External loopback – All traffic received on the ingress interface is blindly sent out of the egress interface.

  • Internal loopback – All traffic received on the egress interface is blindly injected into the ingress interface.

• In the case of Layer 3 interfaces, only external loopback is supported.

• When a Bundle interface is placed into loopback, traffic on all bundle link members are looped back.

• MAC address must always be swapped on looped-back traffic.

• Supports dropping of packets received in the non-loopback direction.

• Allows the application of multiple filters to loopback only a subset of traffic received by an interface and only drop the corresponding reverse-direction traffic.

• Provides an option to specify a time period after which the loopback is automatically terminated.

• Supports at least 100 simultaneous loopback sessions across the system.

These are the limitations of Ethernet Data Plane Loopback (EDPL):

• Layer 3 interfaces including pseudowires are not supported in internal EDPL.

• The first generation Cisco ASR 9000 Ethernet Line Cards are not supported.

• Virtual interfaces such as BVI are not supported.

• Filtering based on LLC-OUI is not supported.

• A maximum of 50 simultaneous loopback sessions are supported for each Network Processor on the linecard.

• LAG bundles that are member of Satellite nV interface over bundle inter-chassis link (also known as LAG over LAG bundles) are not supported.

Use the show ethernet oam configuration command to display the values for the Ethernet OAM configuration for a particular interface, or for all interfaces. The following example shows the default values for Ethernet OAM settings:

• Some of these settings are not supported on certain platforms, but the defaults are still reported. On the Cisco CRS-1 Router, the following areas are unsupported:

• Hello interval configuration

• Remote loopback

• Uni-directional link-fault detection

  
  RP/0/RP0/CPU0:router# show ethernet oam configuration 
  Thu Aug  5 22:07:06.870 DST
  GigabitEthernet0/4/0/0:
    Hello interval:                                              1s
    Link monitoring enabled:                                      Y
    Remote loopback enabled:                                      N
    Mib retrieval enabled:                                        N
    Uni-directional link-fault detection enabled:                 N
    Configured mode:                                         Active
    Connection timeout:                                           5
    Symbol period window:                                         0
    Symbol period low threshold:                                  1
    Symbol period high threshold:                              None
    Frame window:                                              1000
    Frame low threshold:                                          1
    Frame high threshold:                                      None
    Frame period window:                                       1000
    Frame period low threshold:                                   1
    Frame period high threshold:                               None
    Frame seconds window:                                     60000
    Frame seconds low threshold:                                  1
    Frame seconds high threshold:                              None
    High threshold action:                                     None
    Link fault action:                                          Log
    Dying gasp action:                                          Log
    Critical event action:                                      Log
    Discovery timeout action:                                   Log
    Capabilities conflict action:                               Log
    Wiring conflict action:                           Error-Disable
    Session up action:                                          Log
    Session down action:                                        Log
    Remote loopback action:                                     Log
    Require remote mode:                                     Ignore
    Require remote MIB retrieval:                                 N
    Require remote loopback support:                              N
    Require remote link monitoring:                               N
  

Caution

Certain SLA configurations can use a large amount of memory which can affect the performance of other features on the router.

Before you configure Ethernet SLA, consider the following guidelines:

• Aggregation—Use of the aggregate none command significantly increases the amount of memory required because each individual measurement is recorded, rather than just counts for each aggregation bin. When you configure aggregation, consider that more bins will require more memory.

• Buckets archive—When you configure the buckets archive command, consider that the more history that is kept, the more memory will be used.

• Measuring two statistics (such as both delay and jitter) will use approximately twice as much memory as measuring one.

• Separate statistics are stored for one-way source-to-destination and destination-to-source measurements, which consumes twice as much memory as storing a single set of round-trip statistics.

• The Cisco CRS Router supports SLA packet of 100 ms and longer. The overall packet rates for SLA is up to 2000 CCMs-per-second sent, and up to 2000 CCMs-per-second received, per card.

• You must define the schedule before you configure SLA probe parameters to send probes for a particular profile. It is recommended to set up the profile—probe, statistics, and schedule before any commit.

Note	When the once keyword is used for 'send burst' ('send burst once' rather than 'send burst every'), it stops the collection of statistics with the packets that cross probe boundaries.

The following procedure provides the steps to configure Ethernet Service Level Agreement (SLA) monitoring at Layer 2.

To configure SLA, perform the following tasks:

When you configure on-demand SLA operations, consider the following guidelines:

• Each MEP supports up to 50 on-demand operations.

• Each card supports up to 250 on-demand operations.

• On-demand Ethernet SLA operations can be run in addition to any other ongoing scheduled SLA operations that you might have configured, and use similar amounts of CPU and router memory. When configuring an on-demand Ethernet SLA operation, you should consider your existing SLA operation configuration and the potential impact of additional packet processing to your normal operations.

• If you do not specify a schedule for the on-demand operation, the probe defaults to running one time beginning two seconds from the execution of the command, and runs for a ten-second duration.

• If you do not specify the statistics for the probe to measure, it defaults to measuring all statistics, inlcuding these statistics by probe type:

  • CFM loopback—Two-way delay and jitter is measured by default.

  • CFM delay measurement—One-way delay and jitter in both directions, in addition to two-way delay and jitter is measured by default.

  • CFM synthetic loss measurement—One-way FLR in both directions is measured by default.

• The default operation mode is synchronous, where progress of the operation is reported to the console and the output of the statistics collection is displayed.

Note	When the once keyword is used for 'send burst' ('send burst once' rather than 'send burst every'), it stops the collection of statistics with the packets that cross probe boundaries.