ASON-WSON Fundamentals Rev9

ASON/WSON Fundamentals

Welcome to the ASON / WSON Fundamentals e-Learning Course.

1

Objectives During this course we are going to focus on: › An introduction to ASON/WSON to ASON/WSON technology › A description of the ASON/WSON the ASON/WSON Control Plane › A description of the ASON/WSON the ASON/WSON Protections ›On completion of this course the participants will be able to: 1.Describe 1. Describe what ASON/WSON what ASON/WSON is and which are its it s main characteristics. 2.Differentiate 2. Differentiate between ASON between ASON and WSON 3.Understand how WSON works and the relevant characteristics in terms of control plane and 3.Understand protecti prot ection on sche schemes mes introduced into the WDM layer l ayer

During this course we are going to focus on: •An introduction to the ASON/WSON technology •A description of the ASON/WSON Control Plane •A description of the ASON / WSON Protections On completion of this course the participants will be able to: – Describe Describe what ASON is and which are its it s main characteristics. – Differentiate Differentiate between ASON and WSON – Understand Understand how WSON works and the relevant characteristics in terms of control plane and protection schemes introduced into the WDM layer

2

Introduction

Let’s introduce ASON / WSON W SON main general concepts.

3

Terminology Optical Domain

MPLS-TP (ITU-T / IETF)

GMPLS (IETF)

ASON (ITU-T)

OTN (ODU Switching) (IETF)

ASON, ASTN, WSON, GMPLS SDH Domain

ASTN SDH (ITU-T)

WSON WDM

WDM Domain

(IETF)

Generalized Multi-P ulti-Protocol Label Switching (IETF) – GMPLS : Generalized – ASON : Automatic Switched Optical Network (ITU-T / General Optical Domain)

ulti-Protocol Label Switching – Transport Profile – MPLS – TP : Multi-P (Optical Data Unit Switching) – OTN (ODU Switching): Optical Transport Network (O – ASTN : Automatic Switched Transport Network (ITU-T / SDH Domain) – WSON : Wavelength Switched Optical Network (IETF / WDM Domain)

Before beginning with the details about the ASON/WSON ASON/WSON technology and according to the fact that it’s not unusual to listen to people speaking about about ASON, ASTN, WSON, GMPLS, as if all these technologies technologies were the same, it seems to be useful to clarify some points about the terminology. The acronym GMPLS, that stands for Generalized Multi-Protocol Label Switching, refers to a suite of protocols developed by IETF to extend the MPLS ideas outside the context of the IP world. The acronym ASON, that stands for Automatic Switched Optical Network, is a Recommendation Recommendation developed by ITU-T that specifies the requirement to apply the GMPLS technology to a generic optical network; in this context, with the term “generic optical network” we refer both to an SDH or WDM network, network, or even to OTN (ODU switching) an MPLS-TP packet network. In the SDH domain, the specific used acronym is ASTN, that stands for Automatic Switched Transport Transport Network, and, also in this case, case, it is a Recommendation Recommendation developed by ITU-T. It’s a framework that represents the ITU-T IT U-T ideas about how the GMPLS technology should be applied to the SDH world (fast rerouting of Virtual Containers). In the WDM domain, instead, the specific used acronym is WSON, that stands for Wavelength Switched Optical Network, and this a draft, not yet a Recommendation, Recommendation, developed by IETF, that represent the IETF and ITU-T vision of how the GMPLS should be applied to the optical part of the WDM world (fast rerouting of wavelength). wavelength).

4

Terminology

ASON (ITU-T)

GMPLS (IETF)

ASTN

WSON

(ITU-T)

(IETF) GMPLS

FRAMEWORKS

Supporting Protocols

ulti-Protocol Label Switching) – GMPLS (Generalized M ulti-P

GOSP GO SPF F - TE

GOSPF-- TE (Generalized Open Shortest Path First for Traffic Engineering) – GOSPF

GRSV GR SVP P - TE

ReSer Vation Protocol for Traffic Engineering) – GRSVP-TE (Generalized Resource ReS – LMP (Link Management Protocol) – OIF-UNI ( Optical Internetworking Forum – User to Network Interface)

LMP OIF-UNI

Why has WSON been developed by IETF and not by ITU-T? After some discussion, the ITU-T decided decided that it would have have been better to avoid the development development of new protocols to support support ASON, ASTN and WSON; they decided to use the already existent IETF GMPLS suite. In the end, for the last born W SON technology, technology, IETF and ITU-T decided to work together on the project; the result is that WSON W SON related documents are published by IETF and not by ITU-T, even if the latter had an active role in the project. The conclusion conclusion is that ASON, ASTN, ASTN, WSON are frameworks that rely on the IETF GMPLS suite. The GMPLS suite is formed by the GMPLS GMPLS protocol itself and a set of supporting protocols: the GOSPF-TE, the RSVP-TE, the LMP and and the OIF-UNI protocols. The GOSPF-TE and the GRSVP-TE are extensions of already existent protocols developed to support the MPLS-TE technology. The LMP and OIF-UNI are new protocols specifically designed designed to support GMPLS.

5

WSON evolution - advantages Advanced Network Resilience

WSON Advantages

Automatic Circuit Provisioning

Reduced Costs

Why do we need WSON? What are the main advantages of the WSON network compared with a traditional network? There are three main advantages: The most important advantage is that, in WSON, new protection schemas allow advanced network resilience mechanisms, that can react, in an automatic way, also in case of multiple failures. The second advantage is the possibility for WSON to realize the automatic circuit provisioning: as a strategic point of evolution, WSON will be able, in the future, to provide circuits on demand, allowing an IP router, for example, to ask for a circuit directly to the WDM node that is at the ingress of the WSON, without passing through the NMS and the human operator. The third advantage is the strategic cost reduction mainly due to the fact that it is possible to share the protection bandwidth among a group of WSON circuits.

6

a Traditional Network

Network Management System (NMS)

Data Communication Network (DCN)

WDM Node Ethernet Node

Ethernet Node

To implement these advantages, WSON is quite different compared with a traditional WDM network. To understand what are these differences, let’s take a look at what happens in a traditional WDM Network. Let’s consider the WDM network in the picture: this network has 6 interconnected WDM nodes. In a traditional WDM network the nodes are not aware of the fact that they are part of a network. For example, the nodes don’t know the topology of t he network and are not able to take any decision about where to put the customer circuits; even the entity circuit has no meaning for a single WDM node; in other words, the nodes are not network aware. Who has the ability to understand how the nodes are interconnected? Who has the intelligence to understand that a list of cross-connections between two termination points is a circuit? In other words, who is the brain of the network? In a non WSON environment the intelligence of the network only resides inside the NMS and the human operator that sits in front of it. The decision taken by NMS and the human operator are sent individually to each node by the DCN. The nodes, passively execute the commands coming from the NMS; they don’t take part in any routing decision; they are not able to interact with the other nodes in order to create or delete circuits.

7

a wson Network




Ethernet Node

In a WSON network the situation is different: the intelligence of the network is now distributed among the nodes, the NMS and the human operator. Specific WSON processes run inside the controller cards of each node, that implement, together with the WSON processes running inside the NMS, the distributed Control Plane, that is the distributed intelligence that manages the customer traffic.

8

Evolution: from a centralized to a distributed control plane

WSON CENTRALIZED

DITRIBUTED

CONTROL PLANE

CONTROL PLANE

So the first point of distinction between a traditional WDM network and WSON is the evolution from a centralized intelligence to a distributed intelligence, therefore the evolution from a centralized control plane to a distributed control plane.

9

Evolution: from a centralized to a distributed control plane

DISTRIBUTED CONTROL PLANE

DATA PLANE

The distributed control plane controls the data plane, that is the logical entity formed by all the resources available in the network to transport customer traffic. The data plane transports the customer traffic; the control plane decides where this traffic should pass through and sends the right commands to the involved data plane resources. The distinction between Control Plane and Data Plane is important in a W SON context. Please notice that, the fact the control plane is distributed among different entities, means that these entities must be coordinated: the complexity of the WSON Control Plane is mainly due to its distributed nature.

10

WSON internodes Communication network




WSON Inter-node Communication Network Ethernet Node

In order to be able to implement a distributed intelligence, the WSON equipments must have , first of all, the capability to speak with each other, not only with the NMS. This is obtained by an inter-node communication network that is realized using specific WSON reserved communication channels; in GMPLS terminology, these special channels are called Control Channels. It’s important to remark that, this new intra-nodes communication network doesn’t substitute the traditional Data Communication Network, but works together with it. The DCN must be present in any case, to allow the communications between the nodes and the NMS for both WSON and non WSON related operations. For example, the performance data collection is realized using the traditional DCN; the control channels are not involved in this process.

11


IN-FIBER-IN-BAND IN-FIBER-OUT-OF-BAND

Logical Control Channel that runs inside the fiber together with the customer traffic

Depending on the physical implementation it’s possible to classify the control channels in different categories. An in-fiber-in-band control channel is a logical communication channel that shares the bandwidth with the customer traffic and that runs in a fiber. The bandwidth available for the communication is taken by the bandwidth available for the customer traffic; if there is no bandwidth consumption for the communication, it means that more bandwidth is available for the customer traffic. Typically, the routers in the IP world use in-fiber-in-band communication: control information is sent together with the customer traffic. In in-fiber-out-of-band control channels, the bandwidth reserved for the traffic and the bandwidth reserved for the communication are completely not overlapping.

12


OUT-OF-FIBER-OUT-OF-BAND Ethernet INTERFACE

Ethernet INTERFACE

LOGICAL CONTROL CHANNELS IMPLEMENTED USING THE ETHERNET Q INTERFACE (NORMALLY RESERVED FOR THE COMMUNICATION WITH THE NMS)

The other type of control channel is the out-of-fiber-out-of-band control channel: the typical example is when we use the Ethernet interface, that is normally reserved for the communication with NMS, to transport the logical control channel; in this case, the channel is out of fiber, because it runs outside the fiber transporting the customer traffic and it is out of band, because the bandwidth on the Ethernet interface is reserved for the communication and cannot be used to transport traffic.

13


λ1 λ2 λ3 . .. λ80 Optical Supervisory Channel (OSC)

Bandwidth reserved for the in-fiber-out-of-band control channel

As far as DWDM systems are concerned, in particular, the relevant standards define an additional low rate lambda that is specifically designed to transport management information. The logical channel that uses this special lambda is called Optical Supervisory Channel or OSC. Inside the OSC, a configurable amount of bandwidth can be reserved for W SON internodes communication. This type of control channel is in-fiber, because it shares the physical fiber with the customer traffic, and it is also out-of-band, because the amount of bandwidth reserved for the control channel inside the OSC can’t be used to transport any other customer traffic. In special applications, like the single hop long distance, this OSC can also be implemented on one of the wavelength part of the standard grid for working traffic, loosing therefore a traffic channel, but achieving longer distance coverage for the OSC itself.

14

Requirements For a WSON

The nodes must be able to exchange information about the status of the network

What are the general

The nodes must be able to send commands each other

The nodes must be able to re-route the traffic from one line to another line even changing the lambda of the circuit

requirements for a WSON

?

But, what are the general requirements to realize a WSON able to take advantage of the advanced new protection schemas? A WSON is based on the following assumptions: •The nodes are able to exchange information about the status of the network; •The nodes are able to send commands to each other; •The nodes are able to re-route the traffic from one line to another line even changing the lambda of the circuit.

15


ROADM L I N E 3

LINE 4

E N I L 5

LINE 2

LINE 6

LINE 1

LINE 7

LINE 9

L I N E 8

CUSTOMER TRAFFIC

A WDM node that is able to reconfigure the direction of the traffic via software commands is called multi-directional Reconfigurable Optical Add and Drop Multiplexer or ROADM: all WSON capable nodes are multi-directional ROADMs. But how is the capability to change the lambda used by a circuit using only one transponder obtained? This is obtained using a special kind of transponder that is called tunable transponder: this transponder can change the lambda of the circuit changing its internal configuration via software. Please notice that, it’s always the controller of the WSON node, guided by the distributed control plane processes, that triggers the change of direction and, if necessary, lambda for a circuit.

16


THE WHERE DECISION BE TAKEN TOCAN RE-ROUTE BY THE TRAFFIC ?

NMS

WSON NODE

OR

PCE

PCE

One of the WSON requirements is that the WSON nodes should be able to reroute traffic when a failure occurs. In order to be able to create or re-route WSON circuits, the equipment must know which is the best path to choose (click); How is this knowledge obtained? We can have two alternative approaches: The decision is taken by some process running inside the NMS; The NMS tells to the nodes which is the best path to create; The decision is taken by the WSON node itself; it’s able to calculate where to route o re-route the traffic without passing through the NMS. In both cases, the engine that is in charge to calculate the best worker and all the best protections path, according to some specific constraints, that it’s possible to explicitly define, is generically called Path Computation Engine or PCE.

17

The Control Plane

Let’s start describing in deeper details the main features of a WSON Control Plane.

18

Ideal situation 1.

INFORMATION ABOUT

THE NETWORK TOPOLOGY: GOSPF-TE

2.

CALCULATION OF THE BEST PATHS: PCE

WSON NODE 3.

SENDS COMMANDS TO OTHER NODES

TO CREATE AND PROTECT THE CIRCUITS: GRSVP-TE

Let’s now consider an ideal situation for a while. The ideal situation is the one in which the PCE runs inside each single WSON node: this would allow each node to take real-time decision very fast, independently from other nodes and from the NMS. As a consequence, more efficient protection schemas could be implemented. To make it work, the first step is to allow the WSON nodes to know the topology of the entire WSON network. This information can be acquired using one of the GMPLS supporting protocols, the GOSPF-TE. GOSPF-TE is a routing protocol and its purpose is to allow WSON nodes to exchange information about the network topology. This information is: the administrative costs of the links all around the network, that are assigned via configuration; the bandwidth available or the bandwidth used for each link of the network; the status of each link: is the link up or is it down?; other additional administrative information that could have been assigned to the links for different purposes.

19

Ideal situation 1.

IN FOR MA TIO N AB OU T


2.


WSON NODE 3.



The second step is to calculate the best path: having the picture of the network, every WSON node is able to find out what is the best path to f ollow to create or to protect a circuit; the tool that is in charge of calculating best paths is the PCE.

20

Ideal situation 1.

INFORMATION ABOUT


2.


WSON NODE 3.



Step number 3 is the actual creation or re-routing of a WSON circuit: once decided the best path, a WSON node will send requests to the other nodes on the selected path in order to create, re-route or release WSON circuits; this is done using another GMPLS supporting protocol: the GRSVP-TE. GRSVP-TE is a signaling protocol and its purpose is to allow WSON nodes to send commands, and receive feedbacks from the other nodes, in order to create, re-route or delete WSON circuits. Notice that WSON nodes always exchange topology information, using the GOSPF-TE, even if there are no active circuits in the network; On the other side, the signaling protocol, the GRSVP-TE, is active only at the moment when a circuit is created, re-routed or deleted and only between the involved nodes.

21

THE Path computation engine CALCULATE THE BEST PATHS

PERFORM THE RWA

PCE

VALIDATE THE PATHS

Probably, the path computation engine is the most critical component of a WSON. In general the functions of a PCE are: Calculating the best paths; Performing the Routing and Wavelength assignment; Validating the paths.

22


PERFORM THE RWA

PCE

VALIDATE THE PATHS

In a WSON, the work of the PCE is complicated by the fact that calculating the best path in the photonic world is not as easy as it would be in SDH world. In the photonic world there are some additional physical constraints to be taken into consideration, such as Polarization Mode Dispersion, non-linear effects, not perfect signal amplification and so on; these constraints have a more critical impact in WDM world compared with SDH world. For example, in the photonic world, the fact that it’s possible to create a circuit between a node A and a node B and that it’s possible to create a circuit between node B and a node C, doesn’t imply that it’s always possible to create a circuit between node A and node C. Another critical point is that, in WDM world, given the fact that it is not possible to change the frequency of a circuit without going in the electrical domain, it’s necessary to try to minimize the usage of the lambdas all around the network; the problem to find out the best path in terms of administrative cost, minimizing, at the same time, the number of lambdas used in the network is known in literature as the Routing and Wavelength Assignment problem (RWA). The RWA has been demonstrated to be an hard problem to solve; practically, a real PCE will solve the problem using approximated solutions, based on heuristic, that are techniques designed to solve a problem that ignores whether the solution can be proven to be correct, but which usually produces a good solution. In conclusion, the work of the PCE is to resolve, for each circuit that it’s necessary to create o protect, the RWA problem, taking into account all the possible physical impairments specific for the network: every path must be calculated and validated, checking if the path is physically feasible despite all the physical impairments.

23


PERFORM THE RWA

PCE

VALIDATE THE PATHS

It’s not easy to realize an efficient engine that is able to perform all these tasks: especially the validation step can be a long process in a big network. This is the reason why it’s difficult to have PCE running inside the single WSON node. More commonly, WSON nodes interact with a PCE that works off-line, to have the time to plan all the paths performing calculation, RWA and validation. The circuits that pre-calculate and pre-validate the protection paths before a failure happens are called pre-planned protected circuits. In a first phase, all the WSON protections are pre-planned; In a second phase, when the PCEs will run directly inside the WSON nodes controllers, other kinds of protection will be available, like the so called on the fly protections, in which the protection path is calculated and validated on the fly, real-time, only when a failure occurs. Notice that, even in case in which the PCE works off-line, the routing protocol is still required to be running to monitor the real time status of the network resources; otherwise, the nodes would never know if a link that is part of a precalculated protection path is still available or not.

24

acquiring topology information

COST 10

COST 5

COST 20

COST 15

COST 10

COST 20

COST 30

Control Plane configuration: costs assignment

When setting-up a WSON, it’s always necessary to configure some parameters that will be used by the GMPLS processes running in each WSON node; this phase is known as configuration of the control plane parameters. During the configuration of the control plane, one of the steps is to assign some administrative constraint to each link on the network. This operation can be done manually, or using some automatic tool that can help to avoid mistakes. Once configured, these parameters will be used by the GMPLS routing protocol, the GOSPF-TE. Now, let’s see and example of how W SON nodes acquire information about the topology of the network, that is the first step to calculate the best path. One of a parameter to assign to the links is the administrative cost; it represents the cost to send traffic out of the interface connected to that link. This information will be used by the PCE, because it will try to minimize the total cost of a path represented by the sum of the cost of the links it passes through.

25

acquiring topology information LSU

1

2

Information packet

3

LINK 1

LINK 2 LSU

LSU

L S U

COST 10

LINK 7

L S U

• I’m 5the node 1 LINK

COST 5

L S U

COST 20

COST 15

LINK 3

COST 20

• I’m connected to the node 2 COST 10

through link 1,

LSU

the cost of which is 10; LINK 4

LINK 6

4

COST 30

LSU

5

• I’m connected to the node 4

6

through link 7, the cost of which is 5

How does it work? Every node in the WSON is responsible for sending detailed information about its directly connected link to all the other nodes in the network. Consider node 1 in the picture: when the GOSPF-TE process is enabled, after an initial handshake to its adjacent neighbors, the node 1 builds a packet containing at least the f ollowing information in order to define some common parameters: I’m the node 1, I’m directly connected with node 2 through the link 1, that has the cost of 10 and I’m connected to the node 4, through the link 7, the cost of which is 5. In the context of GOSPF-TE, this information packet is called LSU, Link State Update and the piece of information describing each link is called Link State Advertisement or LSA. Once this information packet is created, it’s sent out to all interfaces of node 1. When the node 2 and the node 4 receive the packet coming form node 1, they store the information inside a local topology database and send an updated copy of the packet to all their own adjacent neighbors, but not the one from which the information packet has been received. When the node 3 and the node 5 receive the packet coming from node 2 and node 4, respectively, they store the information inside a local topology database and send an updated copy of the packet to all their own adjacent neighbors, but not the one from which the information packet has been received. Finally, also the node 6 receives the information coming from the node 1 and stores this information inside its local database. The final result is that the piece of network described by node 1, regarding how link 1 and link 7 are interconnected to the other nodes, is known by all the nodes in the network. In a similar way, also the other nodes in the network create an information packet regarding their own directly attached links, and flood it to all the other nodes in the network. At the end of this flooding process, every node will have exactly the same picture of the network. On the base of this picture, the PCE will apply its algorithm to calculate the best paths.

26

acquiring topology information

1

2

3

LINK 1

LINK 2

LSR

LSR

LSR

COST 10

LINK 7

COST 5

COST 20

LINK 5

COST 15

COST 10

LSR

COST 20

COST 30

LSR

LSR LINK 4

LINK 6

4

LINK 3

5

6

In a real network, the identifier of a node, that in GMPLS terminology is called Label Switching Router or LSR, is an IPv4 address, that uniquely identifies the node in the WSON: this is known as the router-id of the node. Also the link identifiers are IPv4 addresses; every link is a point to point link and it’s seen, by the GMPLS protocols, like an IPv4 sub-network containing only two interfaces: the two interfaces of the two node at the edge of the point to point link. All these IPv4 addresses are assigned to the WSON nodes during the configuration phase of the GMPLS control plane.

27

GOSPF-TE: TE information

› Additional Traffic Engineering information:

In GMPLS context, with the term Traffic Engineering, we refer to the capability to take routing decision based on some additional information that is not possible to express only with a fixed cost given to a link. What we have seen till now, regards only the part of the GOSPF-TE that is not related to the traffic engineering functionalities. To support the traffic engineering functionalities, the LSU packet must transport some additional information regarding the links, not only the administrative cost. For example, we can think of assigning an higher cost to links that have a very high percentage of used bandwidth, so we can imagine a dynamic cost associated to the link that can vary with the bandwidth occupancy. Every time that there is a variation on the bandwidth, for example, when a new circuit is created or is released, new LSU are sent by the nodes that are adjacent to the links involved in the paths to all the other nodes. This functionality is difficult to use when the PCE works off-line, because all the path calculation is done in advance and so the dynamic cost cannot be taken in account. This functionality is important when the PCE run directly inside the nodes and can calculate and validate the paths real time. Notice that, even if the PCE works off-line the status of the network must be monitored because the WSON nodes must always know which of the links are available and which are not.

28


› Additional Traffic Engineering information: › Total Reserve-able Bandwidth › Bandwidth Available

The total reserve-able bandwidth: for each link, how many channels can I reserve for WSON paths? All? Only a part? Is this an 80 channels link o a 40 channels link? The Bandwidth available: how many lambda are free in a moment to create new paths?

29


› Additional Traffic Engineering information: › Colors › Shared Risk Link Groups (SRLGs)

Two optional additional information can be transported by the LSU: Colors Shared Risk Link Groups (SRLGs) A color is an attribute that can be assigned to a link, so to distinguish a group of links from the others. For example, a red color can be assigned to a group of military links: when we create a circuit, it’ possible to request that the circuit pass only through the links that have the red color as attribute configured during the control plane configuration phase. If it’s not possible to satisfy the color constraint, the circuit in not created at all. Among the links respecting the color constraint, the one with the lowest cost path is selected. Notice that the color constraint, not only is stronger than the administrative cost constraint, but it is also blocking: if there is no way to connect the two termination point of a circuit, the circuit is not created at all and a message is sent to the operator.

30


› Additional Traffic Engineering information: › Colors › Shared Risk Link Groups (SRLGs)

The concept of Shared Risk Link Group in used by the PCE when calculating the protection path for a circuit. It represents a group of links that, somehow, share some risk. For example two links passing through the same physical conduit in some point share a risk, in the sense that, if someone damages the conduits, probably the two links will be broken at the same time. During the configuration of the control plane is possible to specify the SRLGs of which a link is part of: it’s an additional topological information that can’t be given only with the administrative cost metric. When are SRLGs used? An operator can specify that he would like to have the worker and protection paths passing through completely disjoint SRLGS. If, for example, the worker path pass through links that are part of SRLG 1, SRLG 2 and SRLG 3, the PCE will try to avoid the usage of links that are parte of SRLG1, SRLG 2 or SRLG 3. As it was for the colors constraint, the SRLGs constraint is stronger than the Administrative cost. But, unlike the case of the colors constraint, the SRLG constraint is not blocking: even if it is not possible to respect the SRLGs diversity between worker and protection paths, the circuit is protected, if a path is available.

31

GOSPF-TE: colors NODE 1

LINK 7

COST 5

LINK 1 COLOR RED

LINK 2 COLOR GREEN

COST 10

COST 20

LINK 5 COLOR RED

COST 10 LINK 6

COST 15

LINK 3

COST 20

LINK 4 COST 30 COLORS RED AND GREEN

NODE 4

Control Plane configuration: colors assignment

Let’s now see an example for colors. Consider the network in the picture: during the configuration phase of the control plane, a color attribute has been assigned to link 1, link 2, link 4 and link 5. To link 1 has been assigned the color red; to link 2 the color green; to link 4 two colors: red and green; to link 5 the color red. The attribute color red can represent, for example, links with good Bit Error Rate performance.

32

GOSPF-TE: colors NO PROTECTION PATH AVAILABLE: THE CIRCUIT IS NOT PROTECTED NODE 1

LINK 7

COST 5

LINK 1 COLOR RED

LINK 2 COLOR GREEN

COST 10

COST 20

LINK 5 COLOR RED

COST 10 LINK 6

COST 15

LINK 3

COST 20

LINK 4 COST 30 COLORS RED AND GREEN

NODE 4

Control Plane configuration: colors assignment Failure on a link

The PCE, before calculating the best path, deletes f rom the picture of the network the links not respecting the color red constraint: link 2, link, 3, link 6 and link 7 are deleted from the topology. Now the PCE calculate the best path taking into account the administrative costs; there is only one path available: the one passing through link 1, link 4 and link 5; node 1 sends commands to the nodes involved in the path and the circuit is created. Notice that, once a color attribute has been assigned to the links, it can be used or not; if the operator doesn’t specify a color during a creation of the circuit, the color attribute is not taken into consideration by the PCE. Another point is that, in case the operator wanted this circuit to be protected by another path passing only through red links, if there is a failure on one of the links of the working side, it’s not possible to protect the circuit anymore, because there is no other path that can join node 1 and node 4 passing only through red links.

33

GOSPF-TE: srlgs COST 10/SRLG 1/LINK 1

COST 10/SRLG 1/LINK 2

SRLG 5

SRLG 3

COST 5

LINK 9



SRLG 4

COST 15

LINK 7

COST 10 SRLG 6/ LINK 8

COST 20

LINK 5


Control Plane configuration: SRLGs assignment

Let’s now see an example for SRLGs. As we said, it can be used to inform the PCE that two links pass through the same physical conduit in a point of the network; this information can be useful, because we can suppose that, probably, if there is a damage on the conduit, both the links will go down. When is this information used? An operator can request to the PCE to satisfy an SRLG diversity between worker and protection path during the creation of a circuit.

34

GOSPF-TE: srlgs THE CIRCUIT IS PROTECTED EVEN IF THE CIRCUIT IS PROTECTED TRYING IT IS NOT POSSIBLE TO RESPECT TO RESPECT THE SRLG DIVERSITY THE SRLG DIVERSITY

Failure on a link



SRLG 5

SRLG 3

COST 5

LINK 9



SRLG 4

COST 15

LINK 7


COST 20

LINK 5

COST 30 SRLG 7/ LINK 6 Failure on a second link

Control Plane configuration: SRLGs assignment

Let’s take a look at the picture; Suppose the worker path of a circuit pass through links 1, 3 and 5, that are part of SRLG 1, SRLG 2 and SRLG 4, respectively. If the operator specifies that he wants to have SRLG diversity between worker and protection, this means that the PCE will try to avoid the usage, for the protection path, of links that are part of SRLG 1, 2 and 4. This is done by the PCE, assigning a very high cost to links not respecting the diversity during the calculation of the best path. Anyway, this constraint is not blocking, like it would have been with colors: if there is a path available, even not respecting the diversity, the circuit is protected in any case; in this situation, the cost of the protection path will be very high. Let’s see an example: If there is a failure on link 3, the traff ic is switched on the protection path passing through the SRLG 5, 6 and 7: the SRLG diversity requirement is respected. Now, suppose there is a second failure on the link 6. The traffic is switched on a second protection path passing through SRLG 1, 2 and 4: the SRLG diversity requirement can’t be respected but, anyway, the traffic is protected.

35

GRSVP-TE: signaling a path NMS CIRCUIT REQUEST WITH PATH SPECIFICATION

VALIDATED LSPs OFF-LINE PCE

WSON CIRCUIT REQUEST

Once the PCE has calculated and validated a circuit, worker and protection paths, the next step will be to setup the circuits, that, in a WSON, must always be bidirectional. The validates paths are therefore sent by the PCE to the NMS, that passes these information on to the ingress node for the specific circuit.

36

GRSVP-TE: signaling a path PCE

NODE 2

NODE 1 Control Channel


LINK 1

LINK 7

LINK 2

LINK 5

LINK 3

LINK 6

NODE 6

C o n t r o l C h a n n e l

LINK 4

NODE 5

NODE 4

The management of WSON circuits is delegated to the GMPLS signaling protocol: the GRSVP-TE. GRSVP-TE is used to setup and tear down circuits. A circuit is the entity that is formed by both worker and protection paths. In GMPLS terminology, a path is called Label Switched Path or LSP, because, during the setup, the nodes exchange messages about resource reservations identified by labels. A circuit is formed by one or more LSPs. But what is a label? A GMPLS label represent a traffic resource on a link: there is one label for each lambda that is possible to reserve in a link; for example, if , after the Routing and Wavelength Assignment phase, the PCE has decided to reserve the first channel of the eighty available for a circuit, it indicates this channel with a label, that represents the channel 1. The information about the chosen label is transported inside GRSVP-TE messages.

37

GRSVP-TE: signaling a path PCE

NODE 2


Control Channel

LINK 1

LINK 7

NODE 3

LINK 2

LINK 5

LINK 3

LINK 6

NODE 6


LINK 4

NODE 5

NODE 4

Now, let’s consider an example: an operator wants to setup a protected circuit between node 1 and node 4; he inserts his request inside the NMS. At this point two things can happens: 1. if the PCE process runs inside the nodes, the NMS sends its request of circuit creation to the node 1, that represents the ingress node for the circuit; the PCE calculates the best path for both worker and protection side, performs Routing and Wavelength Assignment and validates the traffic; the output of the PCE is a list of nodes, links and lambdas for both worker and protection side, that will be used by the GRSP-TE to signal the circuit. 2. if the PCE runs off-line, not inside the nodes, the PCE calculates the best path for both worker and protection side, performs Routing and W avelength Assignment and validates the traffic. The output of the PCE is sent to the ingress node, and the GRSVP-TE process will use this information to signal the circuit.

38

GRSVP-TE: signaling a path The circuit is created: the status of the circuit is transmitted to the NMS

PCE

NODE 2

NODE 1 PATH

RSV PATH

Control Channel

RSV

Control Channel

LINK 1

LINK 7

NODE 3

LINK 2

LINK 5

LINK 3

LINK 6

NODE 6

P A T H

LINK 4

NODE 5

R S V


NODE 4

Please, notice that, in both cases, the signaling phase is performed directly by the ingress node, not by the NMS: in the first case, we can think about a centralized routing and a distributed signaling behavior (CD); in the second case, we can think about a distributed routing and a distributed signaling behavior (DD). Now, suppose that the output of the PCE were: node 1 - link1 - node 2 - link 2 node 3 - link 3 - node 4 – lambda 1. The ingress node prepares a GRSVP-TE message, that is called Path message, and sends it t o next node in the list; in the example, node 1 sends a path message to node 2. The Node 2 checks the lambda 1 on the link 2: if it’s still available, it forwards the request to the next node in the list, that is, the node 3; otherwise, it sends a path error message to the node 1 to inform that is not possible to satisfy the request. If all is ok, the node 3 performs the same operations as the node 2 and forward the path message to the node 4. The node 4 recognizes that it’s the egress node of the LSP, and, if all it’s ok, builds a new message, the reservation message (RSV message) to send backward to the ingress node, the node 1. When the reservation message reaches the node 1, the LSP is created and the information is sent to the NMS to align it about the status of the circuit. Optionally, the ingress node can send a confirm message to the egress node to inform it that all the LSP creation operation has been completed successfully.

39

Other GMPLS related concepts

› Other GMPLS concepts:

› Link Component (LC) › Link Cluster (LK) › Control Channel (CC)

Related to the GMPLS technology, there are some other important concepts: the concepts of link component or LC; the concepts of link cluster or LK; the concept of control channel or CC.

40

Link Component

Link Component (LC)

What is a Link Component? In GMPLS terminology a link component represents, in the control plane, a physical link. A GMPLS Control Plane can use a physical bi-directional link if and only if it has been declared as a link component; otherwise, for the W SON control plane, that link doesn’t exist. In the picture, two WSON nodes are interconnected by a couple of fiber; from a WSON point of view, they are interconnected by a single link component.

41

Link Cluster

LC 1 Link Cluster (LK) LC 2

What is a Link Cluster? ITU-T and IETF were worried about scalability problems that could have risen due to the amount of links that a generic transport network could have. If there are many links in a WSON, more GOSPF-TE LSU messages are required; More LSU messages mean more load on the node controllers and more time for PCEs to find out and validate best paths. The idea has been to reduce the number of advertised links, by grouping a number of physical links together, that is, a number of link components, in a logical entity called link cluster; a link cluster is a cluster of link components.

42

Link Cluster

LC 1 Link Cluster (LK) LC 2

Two link components can be inserted in the same link cluster only if they share some common characteristics: same capacity, same set of SRLGs, colors, administrative cost; same nodes as termination points; same kind of link level protection. It’s always necessary to create the logical entity link cluster, even if only one link component is defined, because, the GOSPF-TE exchange information about link clusters and not about link component. During the control plane configuration phase, the normal procedure is: •Step 1: Define an empty link cluster, assigning to it all the GOSPF-TE parameters, like administrative cost, SRLGs, colors, etc. •Step 2. Define one or more link components; •Step 3: Associate one or more link component to the already created link cluster. When all the link clusters are defined, even the one containing only one link component, the GOSPF-TE process will only take them into consideration. This means that, in general, the nodes all around the network are aware only of the link clusters that are present in the network; The details about what is inside a single link cluster are known only by the nodes that are the termination point of that particular link cluster, through a specific GMPLS protocol called Link Management Protocol or LMP.

43

the lmp protocol

LMP Messages

Control Channel (CC) 1

Control Channel (CC) 2

LMP Messages

The Link Management Protocol of LMP has two major f unctionalities: •Setup and maintenance of the control channels defined between two nodes; •Setup and maintenance of the link cluster defined between two nodes. Normally more than one control channel is defined between two nodes for redundancy and load sharing; the LMP is responsible for managing these control channels, for example, activating one secondary control channel, when the primary is down. The other important task of the LMP is related to link clusters: the correct associations between link components and link clusters is maintained and checked by the LMP. For example, if we try to associate two link components that don’t have common characteristics in a common link cluster, the LMP running between the two nodes will forbid the operation. The LMP protocol is the glue that keeps control channel, link component and link cluster together.

44

Protections

Let’s now analyze the new advanced WSON protections in details.

45

WSON Protections

› There are two different kinds of WSON protections: – The Transponder Sharing Protection or Green Protection – The Safe Optical Sub-Network Connection Protection (Safe OSNCP)

There are two different kinds of WSON protections: •The Transponder Sharing Protection or Green Protection; •The Safe Optical Sub-Network Connection Protection (Safe OSNCP).

46

transponder sharing

› In case of Transponder Sharing Protection, the PCE calculates and validates two or more LSPs: – One worker LSP, that is immediately signaled and activated by the ingress node; – One or more protection LSPs, that the ingress node do not signal and do not activate

immediately; they are maintained inside the database of the ingress node, associated to the specific transponder sharing protected circuit.

The first kind of WSON protection is the transponder sharing protection: one worker LSP is activated and one or more protection LSPs are pre-calculated and pre-validated but not activated until a failure occurs. Only one transponder is necessary for this kind of protection; This type of protection in similar to an On the Fly protection, in which, as soon as a failure occurs, a new protection path is calculated, validated, signaled and activated: the only difference is that, for the transponder sharing protection, the paths calculation and validation is done in advance by the PCE off-line; only the signaling and activation phase are done on the fly, when a failure occurs. This is a flexible protection, because the protection bandwidth can be shared among the protection paths of other WSON circuits. The only disadvantage is the time for switching, not only because it’s necessary to signal the new protection path, and this requires an amount of time that depends on the number of nodes, but mainly because adding a new circuit in a WDM network requires an amount of time for equalization and PMD compensation. PMD compensation can be critical especially for high bit rates, like 40 Gigabit per seconds or 100 Gigabit per seconds. Please notice that, in general, planning a WSON only with transponder sharing protections, allows to save a lot of resources in terms of power consumption: this is the reason why this type of protection is also called Green Protection.

47

transponder sharing Third Protection NODE 1

LSP

NODE 2

Link 1

Second Protection LSP

First

L i n k 9

Link 2

Link 5

Protection Link 8 LSP

Link 7 NODE 6

NODE 3

6 n i k L

Worker Link 3

LSP

Link 4 NODE 5

NODE 4

Let’s consider the transponder sharing example in the picture: Suppose an operator would like to have a circuit from node 1 to node 4 that is protected with a transponder sharing protection: one decision that he has to take is how many protection path he wants the PCE to calculate and validate; let’s, for this case, suppose that the operator decided to have one worker LSP and three protections LSPs. Suppose also that we are in the case where the PCE runs off-line. The operator passes his request to the PCE and the output of the PCE is a list of four LSPs, one worker and three protections, all validated. Now suppose that the PCE has decided the following paths: For the worker: node 1- link 1- node 2- link 2- node 3- link 3- node 4 and the lambda chosen is the green one; For the first protection: node 1- link 9- node 5- link 4- node 4 and the lambda chosen is the red one; For the second protection: node 1- link 8- node 6- link 7- node 5- link 4- node 4 and the lambda chosen is the yellow one; For the third protection: node 1- link 1- node 2- link 5- node 5 – link 4 – node 4 and the lambda chosen is the purple one. Please notice that the protection bandwidth is not reserved until a failure happen; this means that the protection bandwidth can be shared with ot her protection paths of other circuits.

48

transponder sharing

NODE 1

NODE 2

Link 2

Link 1

L i n k 9

NODE 3

Link 5

6 k n i L

Link 3

Link 8

Link 7 NODE 6

Link 4 NODE 5

NODE 4

Now, suppose a failure happens on the link 2; the ingress node, the node 1 in this case, checks if all the links on the path for the first protection LSP are ok; if all are ok, the LSP is signaled and activated; if one of the links is not ok, the ingress node checks the second LSP and, if all the links are ok, this LSP is signaled and activated; over wise, the ingress node goes on checking all the possible protection LSPs until one of these has all the links that are ok. In this case, the first protection LSP is good because all its links are ok. If, after a while, before repairing the link 2, a second failure occurs, on link 9, for example, the third LSP is signaled and activated. If there is an other failure, on link 8, for example, the ingress node is able to react because it knows also a third protection LSP and the traffic can be protected. Of course, in case of a fourth failure, the traffic is lost. This type of protection is called transponder sharing because to implement this protection only one transponder is necessary; if we would like to have the capability not only to reroute the circuit, but also to change the lambda, it will be necessary to use a kind of transponder that has the capability to tune its frequency. It’s common to refer to these transponders as tunable transponders.

49

transponder sharing Multi Direction

Line 1

Directionless

Protection

Colorless

Line 1

Trp sharing

Line 2

Line 9

Now, let’s see an example of a possible internal structure of a WSON node (ROADM). The configuration in the picture shows one card for each line direction; this card is reconfigurable via software and is able to block or to allow to pass a particular lambda; we refer to this type of card as Wavelength Selective Switch, or WSS, card. The WSS card must be supported by a real-time equalizer, that can also be embedded in the WSS card, depending on the particular implementation. In a particular configuration that is called, directionless configuration, a WSS card can be used to re-route a traffic coming from transponder. In the example, we can see a circuit, with a transponder sharing protected circuit using a tunable transponder, that is able to change the frequency to the WDM link via software configuration, guided by the node controller. In the starting situation the circuit uses the orange lambda and passes through the line 1. Now, let’s concentrate on what happens in case of a fault.

50

transponder sharing Multi Direction

Line 1

Directionless

Protection

Colorless

Line 1

Trp sharing

Line 2

Line 9

A failure occurs on the line 1: the controller immediately activates its GRSVP-TE process to signal and to activate the pre-calculated protection LSP passing through the line 9. Please notice that, in this example, we changed the lambda, because the orange lambda on the line 9 has been considered as already used by an other circuit; the change of lambda is obtained using a tunable transponder, in a configuration called color-less, that means, with no fixed lambda. If, at this point, a second failure occurs on the line 9, a second pre-calculated LSP is signaled and activated.

51

Safe OSNCP

› In case of the Safe OSNCP Protection, the PCE calculates and validates three LSPs: – One worker LSP and one protection LSP, that are both immediately signaled and activated

by the ingress node; – One more protection LSP, that the ingress node doesn’t signal and doesn’t activate

immediately; it’s maintained inside the database of the ingress node, associated to the specific Safe OSNCP protected circuit.

The second kind of WSON protection is the Safe OSNCP protection: one worker LSP and one protection LSP are activated and one protection LSP is precalculated and pre-validated but not activated until a failure occurs. Two transponders are necessary for this kind of protection; The traffic is sent in both the worker and protection side at the same time, as it would happen for the traditional OSNCP protection. In case there is a failure on the worker side, the traffic is switched on the protection side in less than 50 milliseconds and the second protection LSP is immediately signaled and activated. Now, in case there is a second failure, this time on the first, original, protection side the traffic is switched on the second protection LSP in less than 50 milliseconds.

52

SAFE OSNCP

NODE 1

NODE 2

Link 1

First Protection LSP

Second

Link 2

L i n k 9

Link 5

Protection Link 8 LSP

Link 7 NODE 6

NODE 3

6 k n i L

Worker Link 3

LSP

Link 4 NODE 5

NODE 4

Let’s consider the example in the picture: Suppose an operator would like to have a circuit from node 1 to node 4 that is protected with a Safe OSNCP protection. Also in this case, suppose that we are in the case where the PCE runs off-line. The operator passes his request to the PCE and the output of the PCE is a list of three LSPs, one worker and two protections, all validated. Now suppose that the PCE has decided the following paths: For the worker LSP: node 1- link 1- node 2- link 2- node 3- link 3- node 4 and the lambda chosen is the green one; For the first protection LSP: node 1- link 8- node 6- link 7- node 5- link 4- node 4 and the lambda chosen is the orange one; For the second protection LSP: node 1- link 9- node 5- link 4- node 4 and the lambda chosen is the red one.

53

Safe OSNCP

NODE 1

Worker

NODE 2

NODE 3

Side Link 1 L i n k 9

Firsttraffic is LOST The Protection

Link 2 The traffic is switched on this side in less than 50 ms Link 5

Link 8

6 k n i L

Link 3

Side

Link 4

Link 7 NODE 6

NODE 5

NODE 4

The traffic switch on this side in less than 50 ms and the precalculated secondary protection LSP is signaled and activated

The output of the PCE is passed through the NMS to the ingress node of the circuit, in this example, the node 1, but only the worker LSP (green) and the first protection LSP (orange) are signaled and activated; the other protection LSP is maintained in the database of the node associated to this particular circuit. Now suppose that a failure happens on the link 2: in less than 50 milliseconds the traffic is switched on the first protection side (orange); in the meanwhile, the secondary protection LSP (red) is signaled and activated. If, after a while, and before repairing the link 2, a second failure happens on link 7: the ingress node reacts switching the traffic on the second protection side in less than 50 milliseconds. If a third failure occurs, for example, on the link 9, t he traffic is lost until one of the LSPs is repaired.

54

Safe OSNCP Multi Direction Line 1

Protection

Colorless

Directionless Line 1

Safe-1+1 OSNCP/OUPSR

Line 2

Line 9

Let’s see an example of what happens inside a WSON node in case of a SAFE OSNCP protection. In the picture, it’s possible to see one circuit formed by two LSPs: the orange line represents the worker LSP and the green line represents the protection LSP; two tunable transponders are necessary for the SAFE OSNCP protection. Suppose that a failure occurs on the line 1: The traffic is switched on the protection side in less than 50 milliseconds, as it would happen for the normal OSNCP protection. The original worker path is deleted; a new pre-calculated protection LSP 2 is signaled and activated passing through the line 2. Now suppose that an other failure happens on line 9. The traffic is switched on the LSP 2 in less than 50 ms. In case a third failure occurs, the circuit is not protected anymore, because only two protection LSPs are pre-calculated by PCE for that specific circuit. In conclusion, the Safe OSNCP protection is very fast compared with the Transponder sharing protection; we pay this speed with more bandwidth to reserve for a specific circuit, because the protection bandwidth cannot be shared with other protection LSPs of other circuits and with an additional transponder that must be present.

55

ASON-WSON Fundamentals Rev9

Recommend Documents