New Timing Distribution Mechanism

TICTOC WG, IETF 71th Philadelphia, USA

draft-ji-tictoc-new-timing-distribution-mechanism-00.txt

Kuiwen Ji (jikuiwen@huawei.com)

Background

Route Technique using for timing distribution

Agenda

Background

• Synchronization is typically distributed from one central office to another using the SONET/SDH signal for optical networks.

• Each node has two synchronization sources

- a primary and secondary source. this provides a degree of protection for the

synchronization network

• As a last line of defense, clock hold-over provides minimum service quality for a given time period.

Typical Master-Slave Synchronization Example

1 1

2 2 2

3 3 3 3 3

4 4 4 Primary

Secondary

Today’s Network

• Synchronization planning and distribution is administered manually base on SSM (G.781) usually.

• SONET/SDH networks are primarily implemented in linear and rings architectures.

• Now with the introduction of the network controlled by GMPLS and synchronous Ethernet, it is probable that the transport architecture will shift from linear/ring to mesh architecture.

• Mesh networks will provide more paths/combinations for synchronization distribution.

Using of SSM

• The SSM (G.781) has been used for a long time in Sync network.

• A synchronization coordinator usually determines how best to implement synchronization to each piece of equipment in the network and configure the priority of reference sources to each.

• We are careful to avoid timing-loops when planning there synchronization networks. Not every bi-directional link can be used even if they are available in the ring.

Limitation of SSM

3

1

2

Source 1

211

1

3

32

2

Source 2

3

1

2

Source 1

211

1

2

2

Source 2

3

1

2

Source 1

21

1

1

2

2

Source 2

: Nodes

: Main timing tracing path

: Backup timing tracing path

#

# : priority of reference sources

A(×) B(√) C(√)

Another simple example

1

2 3

Source 1

4

3

1

1

2

2Source 2 5

4 21

11

2

2

: Nodes

: Main timing tracing path

: Backup timing tracing path

#

# : priority of reference sources

• Clock source 1 is assumed to be a higher priority clock than clock source 2 for this example.

• Probably we can plan the synchronization like this.

Multiple failures

1

2 3

Source 1

4

3

1

1

2

2Source 2 5

4 21

11

2

2

: Nodes

: Main timing tracing path

: Backup timing tracing path

#

# : priority of reference sources

• If the source 1 fails and a failure occurs between node 1 and 2, node 1 will go to holdover.

• Node 1 can get the synchronization from blue link but it can’t use it now.

• We can change the priority of each node to make another configuration of course. But for preventing timing-loop we still can’t use every link bi-direction even if it would be possible to use them.

• The point is that no one configuration is best for every type of possible failure condition. There is still limitation.

X

X

Holdover

Normal mesh network

• How could the synchronization be setup? What’s the best configuration?

• We need to be very careful to avoid timing-loop. Thus, we have to give up many of the bi-directional links.

: Nodes

: Main timing tracing path

: Backup timing tracing path

Source 1

Source 3

Source 2

Background

Route Technique using for timing distribution

Agenda

Information distribution

# : Node

Source 1

Source 2

14

2 3

7

5

68

9

• With the GMPLS control plane, it’s possible to know the network topology and the state and condition of links.

• And the reference source attribution, like priority, quality can be distributed through route protocol OSPF.

• So all nodes know the network topology and which source output to be used and traced as the primary timing source.

Calculating the traceability paths

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

• Each node calculates the timing tracing path to the master clock source based on the topology and the primary source.

• From the root of the primary reference, simple calculating algorithm like Dijkstra can be used to establish a shortest path tree.

• The synchronization distribution algorithm would be like a ‘tree’ structure to prevent timing loop.

Building a timing tree

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

• A ‘ready’ message is sent when timing traceability path is setup and operational.

• Each node will not switch to a new synchronization source until it knows the new synchronization source is ready.

• After a node traces to a new timing source successfully, the node will send a message to the next to show it is ready.

MM

M

Building a timing tree

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

• A ‘ready’ message is sent when timing traceability path is setup and operational.

• Each node will not switch to a new synchronization source until it knows the new synchronization source is ready.

• After a node traces to a new timing source successfully, the node will send a message to the next to show it is ready.

M

M

M

Building a timing tree

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

• A ‘ready’ message is sent when timing traceability path is setup and operational.

• Each node will not switch to a new synchronization source until it knows the new synchronization source is ready.

• After a node traces to a new timing source successfully, the node will send a message to the next to show it is ready.

M

M

Building a timing tree

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

• A ‘ready’ message is sent when timing traceability path is setup and operational.

• Each node will not switch to a new synchronization source until it knows the new synchronization source is ready.

• After a node traces to a new timing source successfully, the node will send a message to the next to show it is ready.

Failure occurs between nodes 3 and 7 disrupting the sync path

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

X

#

A second failure occurs between nodes 6 and 7

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

X

#

X

Source 1 Fails

: Node

Source 1

Source 2

14

2 3

7

5

68

9

: Timing tracing path

#

X

SSM=PRC

Interworking with existing networksBITS 1

: Node which doesn’t use automatic techniques

: Main timing tracing path

: Node which use automatic techniques

: Backup timing tracing path

Source1

SSM=PRC

PRC

SSM=SSU

PRCPRC

SSM=PRC

PRC

DNU

PRCPRC

PRC

BITS 4

BITS 2

BITS 3

1

1 2

1

2

2

1 1

1

1

2

2

2 PRC

• All blue nodes could be viewed as ‘one node’ which use traditional SSM at the boundary to interwork with others.

Benefits

• Can be used in future network like Synchronous Ethernet, 1588 or any network with GMPLS.

• Provide survivability (sync traceability) for multiple failures.

• Possibly ease requirements on clock holdover mode by providing traceability in event of multiple failures (i.e., maintain service quality).

• Easy planning and maintenance. People don’t need to do complex work in Synchronization scheme and configuration.

• More…

Next Step

• Timing distribution is very important for synchronization. Comments from the group are always appreciated

• Working with CCAMP with respect to the GMPLS extensions, which supports this feature.