1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider...

73
1 002, Cisco Systems, Inc. All rights reserved. 002, Cisco Systems, Inc. All rights reserved. 002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Cisco Confidential Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco Systems

Transcript of 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider...

Page 1: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

1© 2002, Cisco Systems, Inc. All rights reserved.© 2002, Cisco Systems, Inc. All rights reserved.© 2002, Cisco Systems, Inc. All rights reserved. Cisco ConfidentialCisco ConfidentialCisco Confidential

Fault Management for Provider Bridges

Ali Sajassi & Norm Finn

June 5, 2003

Cisco Systems

Page 2: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

222

Agenda

• OAM Layering

• OAM Functionality

• OAM Message Types

• Putting the pieces together

• OAM Message Format

Page 3: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

333

Layered Approached Reference Model

Endpoint of the corresponding levels

Connecting Point of the corresponding levels

Ethernet Virtual Connection / Trail

Transport Virtual Connection/Trail

Page 4: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

444

Some Terminology

• Ethernet Trail spans across multiple Ethernet connecting points (e.g., multiple Ethernet switches)

• Transport Trail spans across multiple transport connecting points. Examples of Transport Trails are MPLS PW, ATM VC, FR VC, etc.

Page 5: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

555

OAM Types

CE CEPB PB PB PB

CEs: Customer Equipment

PBs: Provider Bridges serve as Ethernet connecting points (they can also terminate an Ethernet Trail)

Ps: Provider nodes serve as transport connecting point

Ethernet VC / Trail

Ethernet VC / TrailEthernet Link

Ethernet Virtual Connection / Trail

Network A Network B

P P P P P P

Page 6: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

666

OAM Types - Continue

1. Ethernet Service Level (or Ethernet Trail) OAM

• Needed for 802.1ad

2. Transport Level OAM

• Defined by underlying transport – e.g., by IETF if transport is MPLS/IP

3. Ethernet Link OAM

• Ethernet Link OAM is defined by 802.3ah

Page 7: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

777

Ethernet OAM Trails within a Provider’s Network

PB

PB: Provider Bridge

u-PE: User-facing PE

n-PE: Network-facing PE

P: Provider Node

xPE devices have Ethernet switching capability

Ethernet Trail

Ethernet Trail

Network A

PB PB PB

Ethernet Trail

(u-PE) (n-PE) (n-PE) (u-PE)(P) (P)

Page 8: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

888

Ethernet Trail OAM Types

• Two general Types

– End-to-end (e.g., between Edge PBs)

– Segment (between any two PBs along the end-to-end path)

Page 9: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

999

Agenda

• OAM Layering

• OAM Functionality

• OAM Message Types

• Putting the pieces together

• OAM Message Format

Page 10: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

101010

OAM Scope

• Fault Management

• Configuration Management

• Account Management

• Performance Management

• Security Management

Page 11: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

111111

OAM Scope - Continue

• Fault Management

– Fault Notification

– Fault Isolation

– Fault Detection

– Fault Recovery

• Scope is limited to the first three bullets

Page 12: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

121212

Fault Management – ATM Example

• Fault Detection

– Continuity Check for soft failure

• Fault Notification

– Alarm Indication Signal (AIS) and Remote Defect Indication (RDI)

• Fault Isolation (Connectivity Verification)

– Loopback test

• Fault Recovery

– Automatic Protection Switching (APS)

Page 13: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

131313

Fault Detection: ATM Continuity Check

• Tx endpoint sends a cell periodically

• Rx endpoint can distinguish between an idle connection and a failed connection

• Rx node generates an RDI in the upstream if it doesn’t receive CC within a time interval

End point 1

ATM CC

Node 1

Node 4

End point 2

Node 2

Node 3

End point 1

ATM CC

Node 1

Node 4

End point 2

Node 2

Node 3

RDI

Page 14: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

141414

Fault Notification: ATM AIS + RDI

Node 1

Node 4

Node 2

Node 3

Failure-A

Failure-B

Downstream from A

Upstream from A

AIS-A

RDI-A

RDI-B

AIS-B

Node 1

Node 4

Node 2

Node 3

Failure-A

Downstream from A

Upstream from A

AIS-A

RDI-A

• The downstream node adjacent to the failure generates an AIS signal to indicate a failure in the upstream

• The receiving endpoint generates an RDI signal for the upstream endpoint (this is an end-to-end OAM message)

Page 15: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

151515

Fault Isolation: ATM Loopback Test

• Tx endpoint sends either a segment or end-to-end Loopback cell

• Rx connecting points (that are not loopback destination) extract the cell and forwrads them in segment loopback

•Rx connecting points relay the OAM cell transparently in end-to-end OAM

• Rx endpoint extracts and process the OAM cell and after modifying the cell, it sends a response in the opposite direction

Page 16: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

161616

Ethernet Trail OAM

• Objectives

– To follow proper layering structure

– To avoid duplicate functionality in other layers

– To have minimum impact to existing IEEE standard

Page 17: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

171717

Ethernet Service Level (Trail) OAM

• Fault Detection

– Continuity Check to detect soft failure (mis-config and software failure) as well as hard failure (node or link failure)

• Fault Notification

– AIS/RDI ?

• Fault Isolation

– Loopback & Traceroute

• Fault Recovery

– Spanning Tree Protocol (STP)

Page 18: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

181818

Fault Detection – Continuity Check

Agg

CE

u-PE

u-PEAggn-PE

n-PE

CE

CE

AggCE n-PE

n-PEAggu-PE

u-PE

CE

Agg u-PE

u-PEAggn-PE

n-PE

CE

CE

Island A

Island C

Island B

LAN Emulation for blue VPLS

• u-PE, Agg, and n-PE are all Provider Bridges (PBs)

• Each u-PE broadcast CC message periodically to all members of a service instance

Page 19: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

191919

Fault Detection – Continuity Check

CE

u-PEAggn-PE CEn-PEAggu-PECE

CC messages

LAN Emulation for blue VPLS

Only generated by u-PE periodically – e.g., once a minute

Page 20: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

202020

Fault Detection – Continuity Check

• Description:– A heartbeat message gets sent periodically from end points similar to ATM CC. In the absence of receiving the heartbeat message for a given interval, the soft/hard failure can get detected

• Operation:– The Edge PB generates an “All Bridge” broadcast message as hearbeat on a per customer service instance basis (e.g., P-VLAN). It is assumed that a PB besides forwarding the message also passes it to its control plane – if not then a special multicast address needs to be used for it.

• Characteristics:– It is only intended for Edge Provider Bridges (E-PB) – e.g., the bridges with UNI for a given customer service instance

– It is used to detect both soft and hard failure

– This message needs to get blocked at the customer UNI

– Relative to loopback function, it is a light-load OAM function and no response is required from the far-end E-PB

– Intermediate nodes use this message to remember the E-PBs’ MAC addresses

Page 21: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

212121

Fault Notification

• For a point-to-point Ethernet connection it is conceivable to have an AIS + RDI similar to ATM case

• But in case of Multipoint Ethernet connection, is there a need for AIS and/or RDI ? If so, what purpose would it serve ?

Page 22: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

222222

Fault Isolation – Loopback test

Agg

CE

u-PE

u-PEAggn-PE

n-PE

CE

CE

Agg

CE

n-PE

n-PEAggu-PE

u-PE

CE

CE

Agg u-PE

u-PEAggn-PE

n-PE

CE

CE

Island A

Island C

Island B

LAN Emulation for blue VPLS

End-to-end loopback

Segment loopback

Page 23: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

232323

Fault Isolation - Loopback

CE

u-PEAggn-PE CEn-PEAggu-PECE

Segment loopback

LAN Emulation for blue VPLS

End-to-end loopback

Page 24: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

242424

Fault Isolation – Loopback test

• Description:– It is used to check the connectivity between two service nodes similar to ATM loopback test

• Operation:– The originating end-point generates a unicast request (aka ping request) and expects a response from the destination node

• Characteristics:– It case be both Segment and End-to-End

– Loopback mode cannot be “hard” loopback (e.g., cross connecting the circuits)

– Only the destination node with the designated MAC DA can respond

– This message needs to get blocked at the customer UNI

Note: 802.2 has defined XID and TEST for loopback test / ping; however, since we need additional FM messages such as TR and CC, we are defining a consistent format for all these messages

Page 25: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

252525

Fault Isolation – Trace-route

Agg

CE

u-PE

u-PEAggn-PE

n-PE

CE

CE

AggCE n-PE

n-PEAggu-PE

u-PE

CE

Agg u-PE

u-PEAggn-PE

n-PE

CE

CE

Island A

Island C

Island B

LAN Emulation for blue VPLS

Page 26: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

262626

Fault Isolation - TraceRoute

CE

u-PEAggn-PE CEn-PEAggu-PECE

Trace Route messages

LAN Emulation for blue VPLS

TR responses

Page 27: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

272727

Fault Isolation – Trace Route

• Description:– It is used to isolate the location of the fault along the end-to-end Ethernet Trail. There is no equivalent function in ATM.

• Operation:– The originating end-point generates a multicast request with the destination MAC DA embedded in it. This message traverses hop by hop. Each intermediate service nodes upon receiving this message, checks to see if it has the MAC DA. If so, it forwards the message to the proper port associated with the MAC DA and respond to the originator. If not, it ignores the message.

• Characteristics:– It case be both Segment and End-to-End

– Only the destination nodes can respond in order to reduce the number of responses

– This message needs to get blocked at the customer UNI

– It can also use multicast/broadcast address as MAC DA in which case, the trace route can used to discover network topology (active nodes)

Page 28: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

282828

Agenda

• OAM Layering

• OAM Functionality

• OAM Message Types

• Putting the pieces together

• OAM Message Format

Page 29: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

292929

OAM Message Types - Three Types

• Simple– OAM message is completely transparent to the intermediate service nodes and simply gets switched by them

• Snoop– OAM message gets snooped by each service node along the path as well as getting forwarded to the next node

• Relay– OAM message gets intercepted by each service node along the path and then gets generated again

Page 30: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

303030

OAM Message Types

CE

u-PEAggn-PE CEn-PEAggu-PECE

Simple

LAN Emulation for blue VPLS

Relay

Snoop

Page 31: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

313131

FM Messages and their types

• Loopback

– Simple

• Continuity Check

– Snoop

• Trace-Route

– Relay

Page 32: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

323232

Loopback: Simple

CE

u-PEAggn-PE CEn-PEAggu-PECE

Loopback or Ping request

LAN Emulation for blue VPLS

Ping Response

Page 33: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

333333

Continuity Check: Snoop

CE

u-PEAggn-PE CEn-PEAggu-PECE

LAN Emulation for blue VPLS

Page 34: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

343434

Trace Route: Relay

CE

u-PEAggn-PE CEn-PEAggu-PECE

TR Relay

LAN Emulation for blue VPLS

Page 35: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

353535

Response Types for TR

• There are two response types for TR messages

– Response from selected nodes – e.g., n-PE only or u-PE + n-PE

– Response from all nodes – e.g., u-PE, Agg, n-PE

Page 36: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

363636

Relay TR Response: Selected v.s. All

CE

u-PEAggn-PE CEn-PEAggu-PECE

RELAY Trace Route

LAN Emulation for blue VPLS

Response from selected nodes

Response from all nodes

Page 37: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

373737

Agenda

• OAM Layering

• OAM Functionality

• OAM Message Types

• Putting the pieces together

• OAM Message Format

Page 38: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

383838

Fault Isolation – is there an issue ?

• Even though CC can be used as periodic refresh of MAC aging timers for u-PEs MAC addresses, upon a failure and network isolation, the u-PEs MAC addresses get aged out from the intermediate nodes.

• Therefore, in order to do proper fault isolation, need to remember u-PEs MAC addresses in the intermediate nodes.

Page 39: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

393939

Steps for Fault Detection & Isolation

1. Issue CC w/ multicast address from u-PEs belonging to a given service instance

2. If CC fails, then issue PING w/ PE MAC address for connectivity verification to the far-end PE (MAC SA of CC msg).

3. If PING fails, then issue Trace Route w/ the target PE MAC address

4. If TR fails and the last egress node is an MPLS/IP node, then issue LSP ping or VCCV to isolate the failure within the Transport Trail

Page 40: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

404040

Failure Management scenario – an Example

Agg

CE

u-PE

u-PEAggn-PE

n-PE

CE

CE

AggCE n-PE

n-PEAggu-PE

u-PE

CE

CE

Agg u-PE

u-PEAggn-PE

n-PE

CE

CE

Island A

Island C

Island B

LAN Emulation for blue VPLS

Page 41: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

414141

Step 1: Continuity Check

• Each u-PE periodically multicasts its heartbeat to other members of the VPLS

• Heartbeat msgs are forwarded by n-PEs and Agg devices but not intercepted

• Heartbeat message (CCs) are of type Snoop

• Heartbeat msgs are terminated by u-PEs and the received u-PE starts a timer for each far-end u-PE

Page 42: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

424242

Step 2: PING

• When a u-PE doesn’t receive the CC (hearbeat) msgs from a remote u-PE (either due to LINK, NODE, or SOFT failure), then it initiates PING toward that remote PE

• PING msg is of type SIMPLE and uses the MAC address of that remote PE

• If PING fails, then operator is notified for further diagnostics (e.g., to run Trace Route)

• If PING passes and CC fails, then this is an abnormal case (e.g., SW corruption) and needs to get reported to the operator

Page 43: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

434343

Step 3: RELAY Trace Route

• If PING fails, then need to locate the failure

• 1st diagnostic test is to run Trace Route

• The last TR response has the cumulative path info and points to the failed node/link

• There are two types of RELAY TR.

– A) response comes from selective nodes (n-PEs and u-PEs)

– B) response comes from every node

Page 44: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

444444

Step 4: Transport Trail test

• If the Relay TR fails, then the failure can be isolated to the last egress PB along the path

• If the connection between last PB and its adjacent downstream PB is virtual (e.g., PW over MPLS/IP), then one can further isolate the failure along the transport trail by using facilities available at the transport level – e.g., in case of MPLS

– use MPLS ping and TR

Page 45: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

454545

Speical Case

• If RELAY TR passes, then there may be a FIB corruption and the operator can initiate multiple loopback tests – one for each pair of u-PE and one of the intermediate service nodes reported in the relay TR.

Page 46: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

464646

Agenda

• OAM Layering

• OAM Functionality

• Putting the pieces together

• OAM Message Types

• OAM Message Format

Page 47: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

474747

OAM Message Format

• The followings are some examples of Ethernet OAM message format for Fault Management messages (e.g., Loopback, Continuity Check, and Trace Route)

• There are some proposal in MEF for Loopback OAM message format (but not for CC and TR since these concepts haven’t been discussed to my knowledge)

• Need to work with MEF and other interested parties in defining and finalizing the format for these messages

Page 48: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

484848

OAM msg format

OAM MAC DA

OAM MAC SA

VLAN TagEther type (VLAN)

Ether type (OAM) Type

OAM Data

Sub-type

FCS

Page 49: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

494949

OAM message format - continue

OAM Type OAM Sub-type

Fault Management Continuity Check

Loopback

Trace Route

Performance Management TBD

Page 50: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

505050

OAM Data: Continuity Check

• No OAM data part has been identified for CC currently

Page 51: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

515151

OAM Data: Trace Route Request

Addr type

Resp. mode

Address of Requesting Service Node

Min. hop to resp

Max hop to resp

Correlation Tag

Requested info bit map

Hop #

Reserved

Piggy-bag Info

Piggy-bag flag

Mac address of the destination Edge-PB

Page 52: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

525252

OAM Data: Trace Route Request

• Hop number

– The hop number of the adjacent upstream node. The receiving node increments this number by one before passing the TR request to the next downstream node.

• Response mode

– Indicates which nodes need to respond

» 01: All nodes need to respond

» 02: Only n-PEs need to respond

» 03: Only the nodes with min. and/or max. hops value need to respond

» etc.

• Min. hop to respond

– Only nodes with hop count greater or equal to this value can respond (a value of zero means there is no minimum hop count)

• Max. hop to respond

– Only nodes with hop count less or equal to this value can respond (a value of zero means there is no maximum hop count)

Page 53: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

535353

OAM Data: Trace Route Request

• Piggy-bag Flag

– Indicates whether piggy bagging of info is allowed or not

• Address type

– indicates what type of address to be used to identify the requesting service node (the same type must also be used in the respond).

» 01: IPv4 (4-byte)

» 02: IPv6

» 03: VPN-id (8-byte)

» etc.

• Address of Service Node

– The address of the service type as specified by the previous field

• Correlation Tag

– The tag used to associated the TR responses to the TR request

Page 54: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

545454

OAM Data: Trace Route Request

• MAC address of the destination Edge-PB

– MAC address of the Edge-PB to which lost connectivity (not receiving CC message from)

• Requested Info Bit Map

– A bit map that indicates what info should be sent in the response

Page 55: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

555555

OAM Data: Trace Route Response

Node Hop #

Node Type

Address of the Requesting Service Node

Correlation Tag

Info from node 1

Info from responding node

EOM

Page 56: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

565656

OAM Data: Trace Route Response

• Node hop #

– Indicates the hop count of the responding node

• Node Type

– Indicates the type of the node

• Address of Requesting Service Node

– Indicates address of requesting service node (copied from the request)

• Correlation Tag

– The tag used to associated the TR responses to the TR request (copied from the TR request)

Page 57: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

575757

OAM Data: Loopback Request

Addr type

Address of Requesting Service Node

Correlation Tag

Reserved

Time Stamp (optional)

Page 58: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

58© 2002, Cisco Systems, Inc. All rights reserved.© 2002, Cisco Systems, Inc. All rights reserved.© 2002, Cisco Systems, Inc. All rights reserved. Cisco ConfidentialCisco ConfidentialCisco Confidential

Appendix

Page 59: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

595959

How to remember Edge-PB MAC addresses

• Two basic approaches

1. Maintain a separate table of E-PBs’ MAC addresses in the software to associate a given E-PB’s MAC address with an egress port.

2. To disable aging timer for E-PBs’ MAC addresses by using a semi-static entries in the forwarding table.

Page 60: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

606060

How to remember Edge-PB MAC addresses – Software Approach

• In the software approach

– Upon a TCN, the primary forwarding table gets flushed; however, the software forwarding table remain intact and its entries maintain the association between E-PBs’ MAC addresses and corresponding egress ports

– Since TR goes hop by hop (through each node’s control plane), the software forwarding table can be used for forwarding TR messages

– When a TCN occurs, the entries of the software table may get obsolete; however, since this table is not used for forwarding of customer data, then it should be O.K. Also upon the next OAM CC message from a E-PB, the corresponding entry for that E-PB gets corrected

– When a failure occurs without a TCN, then the entries of the software has the correct value which can be used for tracing. The entries of the primary table may have already been aged out.

– In this approach the E-PBs can be identified without the use of special MAC addresses

Page 61: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

616161

How to remember Edge-PB MAC addresses – “Hardware” Approach

• In the “Hardware” approach

– A special MAC address needs to be used for each E-PB for the purpose of OAM management

– The aging of these MAC addresses are disabled through the use of static value in the primary forwarding table

– When a CC is received, the corresponding entry for the E-PB’s MAC address of that CC gets updated

– Upon a TCN, all others MAC entries get flushed except these special MAC addresses. However, these special MAC will point to the correct egress ports upon receiving subsequent CC messages

– Since these special MAC addresses are just used for Fault Management propose, it should not impact any other traffic flow

– When a failure occurs without a TCN, then these static entries point to the correct egress ports which can be used in tracing

Page 62: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

626262

Hardware approach - continue

• This is not an issue for the following reasons:

– Only OAM messages get affected and NO customer traffic gets affected at any time

– Next broadcasted CC will correct the MAC entry of a given intermediate node in case of TCN by overwriting the MAC entry w/ correct port-id and thus Rxed CC will be received correctly.

– Continuity Check messages gets broadcasted periodically from each u-PE. So if a failure results in:

» Topology change, then the entry for u-PE MAC address gets overwritten w/ proper port-id upon receiving the next CC msg

» No topology change, then the entries for u-PE MAC address in all the service node remain intact and point toward the failure point

– Given that CC transmission interval is in order of a minute versus topology change (which is in order of seconds), at worst only a single CC message can get lost

Page 63: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

636363

Ethernet Layer-2

En

ds

tation

Brid

ge

Hu

b

Hu

b

Hu

b

Brid

ge

Brid

ge

Hu

b

Hu

b

Ro

uter

L2

L1

Hu

b

Hu

b

EthernetEthernetEthernetEthernet

Layer 2 world (in blue) consists of endstations, e.g. hosts and routers, at the ends, with bridges in the middle, connected by Layer 1 media.

Page 64: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

646464

Proper Layering

Ed

ge

Brid

ge

PE

-rs

Ro

uter

Ro

uter

Ro

uter

Brid

ge

Ro

uter/B

ridg

e

AT

M S

witc

h

AT

M S

witc

h

MT

S

L2

L1

Rep

eater

Rep

eater

ATM10GPseudowireEthernetEthernetover SONET over MPLS RFC1483

• Proper layering allows you to put together a network using very different Layer 1 media.

• Layer 3 tunnel (e.g. EoMPLS) is an emulated Layer 1 (pseudowire) medium in this environment!

Page 65: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

656565

Ethernet Trace Route

L2

L1

RFC1483Ether/SONET

Pseudowireover MPLS

HubbedEthernet over ATM

Ethernet Trace Route shall be independent from underlying physical or emulated links

Ethernet TR shall operate at L2 level

Page 66: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

666666

Mixed Layering

L2

L1

ATMATMRFC1483

ATMATMRFC1483 RFC1483 RFC1483

One could define an ATM OAM (or MPLS OAM cell), for example, that carried a MAC address from bridge to bridge, “popping up” at each one, and re-submerging into ATM to traverse the next VC.

But that would create dependency on L1 and lots of inter-operability issues among different emulated L1 interfaces (e.g., ATM 1483, FR 1490, EoMPLS, EoL2TPv3, EoSONET, etc.)

Page 67: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

676767

Issues

1. How do we differentiate Customer’s OAM from Provider’s OAM frames when same Ether Type is used for both?

– If we do translation of Ether-type, then that may be inconsistent w/ our 802.1ad strategy (802.1ad doesn’t translate Ether Type, it simply adds additional Ether Type)

– If we do insertion of Ether-type, then we will end up inserting double tag in font of customer’s tag !!

2. Easier alternative would be to use different Ether Types for provider’s OAM versus customer OAM

– If customer uses provider’s OAM Ether Type, then drop the packet

– Otherwise the packet is passed through provider’s network as a regular packet and provider’s bridges are oblivious to it.

Page 68: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

686868

Issues – Using SA MAC for Forwarding Decision

• Majority of time DA MAC is used for forwarding decision. However, there are two scenarios where SA MAC can be used:

– Load balancing over a set of 802.3ad links (besides SA MAC, other fields within payload such as IP five-topple can be used for hashing)

– Load balancing over blocked ports (proprietary mechanism)

• In such scenarios, customer frames can take a different path than provider’s OAM frames

Page 69: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

696969

Failure Type

E2E path Failure

Permanent

(e.g., node or link failure without alternative path)

Transient

(e.g., node or link failure with alternative path – causing TCN)

TR pinpoints the failed node/link

(Relay TR can pinpoints the failed nodes at any time because PE MAC doesn’t age out)

TR doesn’t come into picture

(some nodes will point incorrectly but they get corrected upon next periodic OAM CC message. Only OAM CC msgs gets affected for a short period)

Page 70: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

707070

PE MACs versus Customer MACs

• Advantages of PE MACs in OAM msgs:

– fully under control of SP

–Age timers can be disabled

» if a link/node fails w/o re-route, then use of TR w/ PE MAC address can pinpoint the failure location

» if a link/node fails w/ re-route, then MAC address learning mechanism will self correct the FDB entries upon receiving the next message from the source

• Disadvantages of Customer MACs in OAM msgs:

– can timeout quickly and thus making it very difficult to do tracing for fault isolation

» will result in flooding and lots of non-relevant responses

» can’t pinpoint the problem because there is no way to identify the egress node that used to have the customer MAC address

Page 71: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

717171

OAM msgs and MAC addresses

Known Unicast

(PE MAC)

Known Unicast

(Cust. MAC)

Unknown Unicast

(PE MAC)

Unknown Unicast

(Cust. MAC)

Multicast

Continuity Check

N/A N/A N/A N/A • SNOOP

• OAM-type = Heartbeat

• Set MAC SA timer

Ping • SIMPLE • SNOOP ?

• only egress u-PE returns ACK

N/A N/A N/A

Trace Route

• SNOOP

• RELAY

• SNOOP ?

• RELAY ?

N/A N/A • SNOOP

• RELAY

Page 72: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

727272

CC versus PING for fault detection

• CC & PING comparison

– CC requires O(n) msgs; whereas, PING requires O(n**2)

– CC requires no ACKs; whereas, PING requires ACKs

– both CC & PING need to start after VPLS instance initialization in order to populate FDB along the path properly

– PING cannot be started after failure occurs (w/o CC) since most likely there won’t be any ACK and TR won’t be able to pinpoint the problem because of unknown PE MAC addresses

Page 73: 1 © 2002, Cisco Systems, Inc. All rights reserved. Cisco Confidential Fault Management for Provider Bridges Ali Sajassi & Norm Finn June 5, 2003 Cisco.

737373

Special Case-2: Ping

1. If PING passes and CC fails, then there may be SW corruption in egress u-PE. Although SW corruption should typically disable both CC and Ping

2. If OAM ping (w/ PE MAC) passes, and the problem still persists, then one may want to do OAM ping w/ customer MAC. In such case the problem can be

• A) at the egress interfaces in which case, OAM ping w/ customer MAC won’t detect the problem either

• B) w/ FDB corruption; however, checksum on the FIB should catch this

• C) there is a wrong ACL on customer MAC

• D) there is some sort of load balancing (across either 802.3ad set of links or across blocked links – later one is a proprietary method)

3. Does it make sense to use Ping w/ customer MAC since after a few minutes the entries get aged out and so untraceable ?