Data-centric optical networks and their survivability

TLDR: This paper describes how core networks will evolve to optical transport networks (OTNs), which are optimized for the transport of data traffic, resulting in an IP-directly-over-OTN paradigm.

Abstract: The explosive growth of data traffic-for example, due to the popularity of the Internet-poses important emerging network requirements on today's telecommunication networks. This paper describes how core networks will evolve to optical transport networks (OTNs), which are optimized for the transport of data traffic, resulting in an IP-directly-over-OTN paradigm. Special attention is paid to the survivability of such data-centric optical networks. This becomes increasingly crucial since more and more traffic is multiplexed onto a single fiber (e.g., 160/spl times/10 Gb/s), implying that a single cable cut can affect incredible large traffic volumes. In particular, this paper is tackling multilayer survivability problems, since a data-centric optical network consists of at least an IP and optical layer. In practice, this means that the questions "in which layer or layers should survivability be provided?" and "if multiple layers are chosen for this purpose, then how should this functionality in these layers be coordinated?" have to be answered. In addition to a theoretical study, some case studies are presented in order to illustrate the relevance of the described issues and to help in strategic planning decisions. Two case studies are studying the problem from a capacity viewpoint. Another case study presents simulations from a timing/throughput performance viewpoint.

Figures

Figure 6: illustrates how merging of working and backup LSPs is realized. Both incoming LSPs have their own entry in the Label Information Base (LIB) and these entries target the same output interface and label. The router simply forwards any packet coming in through either working or backup LSP.

Figure 16: illustrates the increase of amount of required spare resources, due to dedication, for a set of topologies with different degrees of meshedness. The figure illustrates that this is most drastic for Local Protection, especially on sparse topologies (which are unfortunately typical for optical transport networks).

Figure 17: Simulation topology investigating effect of protection switching on TCP

Figure 10: shows the impact of a node failure on two traffic flows between the client layer nodes a and c. The left side of the figure illustrates that the logical links a-d and d-c (both terminating in node D) cannot be restored by the server layer, resulting in the isolation of client layer node d. This implies that the first flow a-d-c (transitting this isolated node d) cannot be recovered by the server layer, but that the client layer has to recover this flow, as illustrated by the right side of the figure. The second flow is routed over a direct logical link between node a and c. This logical tranists the failing node D and thus can be restored by the server layer recovery scheme.

Figure 20: MPLS topology (left) and OTN topology (right)

Figure 3: illustrates that in the overlay model (top) the client network is controlled by a separate control plane, independent from the control plane of the transport network. This is in contrast to the peer-model (bottom) where the control plane of the client network is integrated into the control plane of the transport network: thus, colocated client and transport network equipment are seen as a single entity.

Full text

Submission for IEEE Journal on Selected Areas in Communications,
special issue on WDM-BASED NETWORK ARCHITECTURES
DATA-CENTRIC OPTICAL NETWORKS AND THEIR
SURVIVABILITY
Didier Colle, Sophie De Maesschalck, Chris Develder, Pim Van Heuven,
Adelbert Groebbens, Jan Cheyns, Ilse Lievens, Mario Pickavet, Paul Lagasse, Piet Demeester
Ghent University - IMEC, Department of Information Technology
Sint-Pietersnieuwstraat 41, 9000 Gent (Belgium)
tel. no. +32 9 267 35 93
fax. no. +32 9 267 35 99
e-mail {didier.colle, sophie.demaesschalck, chris.develder, pim.vanheuven,
adelbert.groebbens, jan.cheyns, ilse.lievens, mario.pickavet, lagasse, demeester}@intec.rug.ac.be
Abstract. The explosive growth of data-traffic, for example due to the popularity of the Internet,
poses important emerging network requirements on today’s telecommunication networks. This
paper describes how core networks will evolve to Optical Transport Networks (OTNs), which are
optimised for the transport of data-traffic, resulting in a IP-directly-over-OTN paradigm.
Special attention is paid to the survivability of such data-centric optical networks. This becomes
more and more crucial, since more and more traffic is multiplexed onto a single fiber (e.g,
160*10Gbps), implying that a single cable cut can affect incredible large traffic volumes. More in
particular, this paper is tackling multi-layer survivability problems, since a data-centric optical
network consists of at least an IP and optical layer. In practice, this means that the questions “in
which layer or layers to provide survivability” and “if multiple layers are chosen for this purpose,
then how to coordinate this functionality in these layers” have to be answered.
In addition to a theoretical study, some case studies are presented in order to illustrate the
relevance of the described issues and to help in strategic planning decisions. Two case studies are
studying the problem from a capacity viewpoint. Another case study presents simulations from a
timing/throughput performance viewpoint.
Keywords: multi-layer survivability, MPλS, MPLS, IP-over-OTN, recovery, capacity
dimensioning.

D. Colle, et al., Data-Centric Optical Networks and their Survivability 2
Submission for IEEE Journal on Selected Areas in Communications,
special issue on WDM-BASED NETWORK ARCHITECTURES
1. Introduction: from IP/ATM/SDH/WDM to IP-MPLS directly
over OTN-MPλS
The popularity of the Internet [1], [2] has lead in recent years to an explosive growth of the traffic to be carried
by telecommunication networks. Since a few years data traffic even dominates voice traffic [3]. and recent
forecasts don’t seem to predict a quick slowdown of this greediness [4] [3].
It is obvious that this will have a major impact on today’s telecommunication networks. These networks will be
more and more optimized for the dominant data (mainly IP) traffic. Today, a typical (core of a)
telecommunication network consists of a transport network carrying the traffic of several parallel services: e.g.,
Plain Old-switched Telephone Service (POTS), leased-line services, etc. Such a Transport Network (TN) may
e.g. consist of an ATM network (functioning as service integration layer) on top of an SDH network. Fiber
exhaust is currently solved by multiplying the capacity of a fiber ten – or even hundred – times by means of
point-to-point Wavelength Division Multiplexing (WDM) systems. Recently, WDM-systems of 160 10Gbps
wavelengths have been announced [5]. This multiplexing technique has proven to be very cost-efficient due to
the economy-of-scale [6].
It is obvious that incumbent operators also want to profit from the new Internet Service Provider (ISP) market
fragment. They are at a more comfortable position, since they still have their important revenue-generating
voice [3] business and other services, in contrast to new-comers. However, they are of course not willing to
immediately replace their current infrastructure and thus they start their ISP business by running their IP
network in parallel with their currently existing network services, on top of the same transport network. This
means they typically are in (or have just left) an IP/ATM/SDH/WDM multi-layer scenario [7]. The practical
meaning of this scenario is explained in Figure 1.

D. Colle, et al., Data-Centric Optical Networks and their Survivability 3
Submission for IEEE Journal on Selected Areas in Communications,
special issue on WDM-BASED NETWORK ARCHITECTURES
IP
ATM
SDH
IP
ATM
SDH
WDM WDM
ATM VCs/VPs:
cell switched
SDH VC-Ns:
fixed bandwidth bitpipes
(e.g., N=4->150Mbps)
STM-N wavelengths
fixed bandwidth
(e.g., N=16 ->2.5G bps)
Fiber
IP packets
Figure 1: illustrates the IP/ATM/SDH/WDM technology mapping. IP routers exchange IP packets, by
sending them through ATM connections, which requires encapsulation of an IP packet in many ATM
cells. ATM nodes are interconnected by fixed bandwidth bitpipes (VC-Ns) through the SDH network.
The capacity on the fibers interconnecting the SDH DXCs is increased by multiplexing multiple
wavelengths onto a single fiber.
The transport of IP packets through ATM has some major drawbacks. First of all, there is the important cell tax:
approximately 10% overhead (5 bytes header per 48 bytes payload). Secondly, an IP packet has a typical length
of 500 or 1500 bytes [8] and is thus typically encapsulated in many ATM-cells. This implies that per IP packet
many ATM cells have to be handled and processed in intermediate ATM nodes. Yet another disadvantage is
that there is an extra layer to maintain and manage. Of course, ATM also has its benefits: its connection-
orientation, opening opportunities for Traffic Engineering (TE), due to the decoupling of routing (control plane)
and forwarding (data plane).
However, the steady and ongoing progress and research in optimizing IP router designs [9] implies that IP
doesn’t have to take the drawbacks of ATM for granted, if it would be able to overcome its lack in TE-
capability. The MultiProtocol Label Switching (MPLS) concept, grown within the IETF, has proven to be
suitable for this purpose [10], [11], [12], [13]. Thus, in the end, we may expect that an MPLS-empowered IP-
network absorbs the TE-feature of ATM and bypasses the ATM-layer, by coding the MPLS-labels in a shim-
header in front of the IP-packet. Similar to ATM, a Label Switched Router (LSR) will label-switch the packets
(i.e., look up the incoming <interface, label>-pair in the Label Information Base (LIB), in order to know along
which interface to forward the packet with which label). This bypasses the legacy cumbersome lookup

D. Colle, et al., Data-Centric Optical Networks and their Survivability 4
Submission for IEEE Journal on Selected Areas in Communications,
special issue on WDM-BASED NETWORK ARCHITECTURES
operation of the destination address in the routing table. To populate the LIB with appropriate mapping
information, a protocol (either the Label Distribution Protocol (LDP [14]) or the Resource reSerVation Protocol
(RSVP [15])) in the MPLS control plane will be used, allowing to setup and tear down so-called Label
Switched Paths (LSPs) through the MPLS network. (Note that in the remainder of this paper we will use the
following terminology, to refer to an “IP” network: IP-network refers to an MPLS-incapable network, MPLS-
network is short for an MPLS-capable IP network, and IP-MPLS network will be used when it can be either an
IP-network or an MPLS-network. It also may happen that we call an MPLS-network an MPLS-
empowered/capable IP-network (to stress the MPLS capability). The services and traffic (demand) carried by an
IP-MPLS network are always indicated by IP-services and IP traffic respectively.)
Even more, the steady growth of the IP traffic (will soon) allow(s) bypassing the ATM-layer, simply because
the SDH switching granularity (will) match(es) the required line-speeds for the direct interconnection of IP-
MPLS routers. IP-MPLS-router interface-cards of up to 622Mbps or even 2.5Gbps are currently commercially
available and deployed [9], [16]. As traffic won’t stop growing, in no time SDH Digital Cross-Connects (SDH
DXCs) won’t be able anymore to catch up with the required switching granularity (a coarse granularity of the
underlying layer is beneficial for the IP-MPLS network from a scalability point of view). At that moment, SDH
will be bypassed as well and the cross-connect functionality will be pushed into the optical domain, resulting in
a so-called Optical Transport Network (OTN). Optical Network Elements (ONEs) with limited flexibility are
already commercially available and full-flexible large Optical Cross-Connects (OXCs) are ready for massive
commercialization [5], [17].
A final consideration in our roadmap for next generation networks is the fact that transport networks tend to be
rather static, due to the fact that an operator has to setup each connection manually through the Network
Management System (NMS). This doesn’t match with the exponentially growing and highly dynamic IP traffic
pattern, requiring frequent changes of the wavelength bandwidth pipes provisioned by the OTN-network to
carry the IP-MPLS network traffic. Therefore, a current hot research topic is to investigate how this
provisioning process can be automated. As in all switched networks, the control plane will serve this need, as
illustrated in Figure 2. Signaling through the control channel of the User-Network Interface (UNI) – thus
between the IP-MPLS and OTN network – (e.g., OIF UNI spec 1.0 [18]) makes it possible for the client to
automatically request the setup of a new lightpath through the OTN. The control channel through the Network-
Network Interface (NNI) allows the exchange of signaling messages for routing protocol information exchange

D. Colle, et al., Data-Centric Optical Networks and their Survivability 5
Submission for IEEE Journal on Selected Areas in Communications,
special issue on WDM-BASED NETWORK ARCHITECTURES
(e.g., Link-State Advertisements (LSAs) being used in the Open Shortest Path First (OSPF) routing protocol),
setup of a lightpath, etc.
Data-plane
Mgmt-plane
MIB
MIBMIB
MIB
MIB
OXC
OXC
OXC OXC
OXC
Network Mgmt System
Network Element
Mgmt Agent
Telecommunication
Mgmt Network
OXC switch fabric
Customer Premise Equipment
Lightpath (circuit)
Control-plane
Data-plane
OXC switch fabric
Customer Premise Equipment
Lightpath (circuit)
OXC
OXC
OXC OXC
OXC
User-Network
Interface
Network-Network
Interface
OXC controller
Control Channel
Figure 2: shows the difference between a static Optical Transport Network (OTN) at the top and an
Automatically Switched Optical Transport Network (ASON) at the bottom of the figure. An ASON is an
OTN, empowered with a (distributed) control plane (taking over a large part of the crucial functionality
of the management plane), allowing signalling with the client through the UNI, in order to realize a
switched optical channel service.
Generally speaking, there exist two main (extreme) models for an automatic switched optical network. ITU-T
G.astn [19] targets an overlay model for an Automatically Switched (Optical) Transport Network (ASTN is a
generalization of ASON). In the overlay-model, both the transport and its client networks have a separated and
independent control plane. The IETF targets more a peer-model with the Generalized-MPLS (G-MPLS)
concept. This concept originated from MPλS, where the idea was that a wavelength (Lambda) is a label as any

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Data-centric optical networks and their survivability" ?

This paper describes how core networks will evolve to Optical Transport Networks ( OTNs ), which are optimised for the transport of data-traffic, resulting in a IP-directly-over-OTN paradigm. More in particular, this paper is tackling multi-layer survivability problems, since a data-centric optical network consists of at least an IP and optical layer. In practice, this means that the questions “ in which layer or layers to provide survivability ” and “ if multiple layers are chosen for this purpose, then how to coordinate this functionality in these layers ” have to be answered. In addition to a theoretical study, some case studies are presented in order to illustrate the relevance of the described issues and to help in strategic planning decisions. 

Q2. What is the main drawback of a hold-off timer?

The main drawback of a hold-off timer is that higher layer recovery actions are always delayed, independent of the failure scenario.• 

Q3. What are the drawbacks of MPLS rerouting?

One of the drawbacks of MPLS rerouting is that it may suffer from similar inefficiencies as the IP routing protocols on which it is relying: e.g., rather long convergence times, temporary instabilities and loops, etc. 

Q4. What is the first solution to a network failure?

3.2.1 Uncoordinated approachA first solution is to deploy a recovery scheme in multiple layers, without any coordination, resulting in parallel recovery actions at distinct layers. 

Q5. What is the main idea of a backup LSP?

the backup LSP consists of two parts: a reverse part, allowing the local protection switch, and a diverse part from the ingress to the egress, in order to get the protected traffic on the backup LSP through the network. 

Q6. What is the disadvantage of a hold-off timer?

A disadvantage, comparedSubmission for IEEE Journal on Selected Areas in Communications,special issue on WDM-BASED NETWORK ARCHITECTURESto a hold-off timer interworking, is that a recovery token signal needs to be incorporated in the standardization of the interface between network layers. 

Q7. What is the advantage of the bottom-up approach?

An advantage of this approach is that a higher layer can more easily differentiate traffic with respect to the service typesSubmission for IEEE Journal on Selected Areas in Communications,special issue on WDM-BASED NETWORK ARCHITECTURESand thus it may try to restore high priority traffic first. 

Q8. What is the main advantage of MPLS rerouting?

The router simply removes the failing equipment from the link-state database and calculates a new route from itself towards the egress LSR: this implies that the part of the LSP upstream from the failure is not rerouted.