Novel network architecture for optical burst transport [Elektronische Ressource] / Christoph Gauger

Informations

• Publié par: universitat_stuttgart
• Publié le: 01 janvier 2006
• Nombre de lectures: 14
• Langue: English
• Poids de l'ouvrage: 2 Mo

Extrait

Communication Networks and Computer Engineering
Report No. 92
Christoph Gauger
Novel Network Architecture for
Optical Burst Transport
Universität Stuttgart
Institut für Kommunikationsnetze und Rechnersysteme
Prof. Dr.-Ing. Dr. h. c. mult. P.J. KühnChristoph Gauger
Novel Network Architecture for Optical Burst Transport
Universität Stuttgart
Institut für Kommunikationsnetze und Rechnersysteme
(Communication Networks and Computer Engineering,
Report No. 92, 2006)
© 2006
Universität Stuttgart
Institut für Kommunikationsnetze und Rechnersysteme
Printed in GermanyNovel Network Architecture for Optical Burst Transport
Von der Fakultät für Informatik, Elektrotechnik und Informationstechnik
der Universität Stuttgart zur Erlangung der Würde
eines Doktor-Ingenieurs (Dr.-Ing.) genehmigte Abhandlung
vorgelegt von
Christoph Martin Gauger
geb. in Esslingen am Neckar
Hauptberichter: Prof. Dr.-Ing. Dr. h. c. mult. Paul J. Kühn
Mitberichter: Prof. Biswanath Mukherjee, Ph.D., UC Davis, USA
Tag der Einreichung: 09. Mai 2006
Tag der mündlichen Prüfung: 09. August 2006
Institut für Kommunikationsnetze und Rechnersysteme
der Universität Stuttgart
2006To Susanne.Summary
Transport networks form the backbone of communication networks by cost-ef ciently offering
huge bandwidth and by guaranteeing a high service quality and availability. These requirements
can best be met by using optical communication technologies. Currently, wavelength-switching
is the most prominent network technology employing optical ber communication and wa-
velength division multiplexing. It transports data in circuit-switched wavelength channels, the
so-called lightpaths. While for years progress in optical networks has been de ned by ever in-
creasing transmission bit-rates, higher exibility and manageability as well as multi-service and
multi-layer integration are equally important criteria today. Accounting for these trends, optical
burst switching (OBS) has been proposed as a highly dynamic optical network architecture. It
offers ne-granular transport of different packet-switched services and applies statistical multi-
plexing directly in the optical layer.
This thesis presents the design, modeling, and evaluation of the optical burst transport net-
work architecture (OBTN). The architecture is motivated by the need for exible, scalable, and
cost-ef cient transport in next generation networks. In addition, it is stimulated by the research
activities towards highly dynamic optical network infrastructures.
OBTN de nes a network architecture to transport and switch optical burst data in a core net-
work. The design objectives for the OBTN architecture are (i) an overall high quality of service,
(ii) a network design allowing for cost-ef ciency and scalability, and (iii) a network evolution
perspective based on the current wavelength-switched networks. These objectives are achieved
by combining selected concepts, architectures, and strategies of optical burst and optical packet
switching as well as of multi-layer traf c engineering.
In order to provide the background information for the design of OBTN, Chapter 2 introduces
the general characteristics, requirements, and trends for next generation transport networks. Al-
so, it discusses the concept of layering in next generation networks and its application in layer
networks for the virtualization of transport resources. Consequently, virtual topology design and
dimensioning are analyzed to quantify the trade-offs regarding connectivity and resource requi-
rements. Chapter 2 also reviews the fundamental technologies as well as currently emerging
data and control plane architectures for optical transport networks. This presentation is then ex-
tended towards a long-term perspective. It describes architectural constraints and classi cation
criteria for highly dynamic optical network architectures. These criteria are used to characterize
the fast optical circuit switching, optical burst switching, and optical packet switching archi-
tectures. Then, hybrid optical network architectures are discussed as a framework to combine
wavelength-switched and optical burst/packet-switched networks.
iii Summary
Chapter 3 discusses the state of research and technology for optical burst switching to struc-
ture the design space and identify promising approaches. Thus, it presents the requirements for
the different functions in an OBS network and classi es the proposed architectures and me-
chanisms. Particularly, it addresses contention resolution which is necessary to achieve a high
QoS in burst-switched networks. Here, wavelength conversion, ber delay line buffering, alter-
native/de ection routing, and their combinations are looked at. It is concluded that wavelength
conversion is a promising primary contention resolution strategy but should be complemented
by FDL buffering and/or alternative routing. Thus, architectures, parameters, and operational
strategies for FDL buffers are discussed in detail. This is supported by Appendix A which
analyzes the performance of shared FDL buffers for different con gurations and traf c cha-
racteristics. The review of alternative/de ection routing shows that it can only support other
contention resolution schemes if it is closely controlled, i.e., if extensive de ections and rou-
te variations are avoided. Finally, architectures and realization aspects for burst-switched core
nodes are presented to understand their resource and scalability constraints.
Chapter 4 presents the design rationale for OBTN and explains how OBTN combines a burst-
switched client layer network with a wavelength-switched server layer network. Then, it in-
troduces its fundamental concepts, namely the dense virtual topology, constrained alternative
routing, and shared over ow capacity. These components are analyzed regarding their conse-
quences for the overall node and network architecture. Further architectural details and variants
as well as operational strategies of OBTN are discussed to complete the presentation. Final-
ly, a qualitative discussion of OBTN with respect to optical burst switching and hybrid optical
networks concludes this chapter.
Chapter 5 describes a uni ed resource model which allows to dimension and evaluate burst-
switched architectures with different virtual topologies. Also, it details the dimensioning pro-
cess for OBTN. Then, it addresses the simulation methodology and the reference evaluation
scenario used in Chapter 6. It discusses metrics for node and network resources as well as for
QoS performance. Finally, it derives QoS objectives for burst-switched core networks.
Chapter 6 evaluates OBTN and compares it with the two burst-switched reference architec-
tures OBS and Burst-over-Circuit-Switching. OBS uses a sparse virtual topology while BoCS
employs a full-mesh virtual topology. The evaluations in Section 6.1 show that for the same
high target QoS, suitable OBTN dimensionings require substantially less resources in burst-
switched nodes than OBS and slightly less than BoCS. This improvement comes at the cost
of higher resource requirements compared to OBS in the underlying wavelength-switched ser-
ver layer. However, applying the cost relations for lambda grid networks, in which bandwidth
is considered a commodity and client layer resources the major cost driver, OBTN yields an
overall cost reduction. The best results for OBTN are obtained when approximately 10 % of the
network capacity is assigned as shared over ow capacity.
The comparison of OBTN and OBS is extended towards OBS architectures without an FDL
buffer and OBS architectures with alternative routing. It is demonstrated that bufferless OBS
with and without alternative routing requires approximately the same amount of server layer
resources as OBTN. However, it consumes more client layer resources. For OBS with an FDL
buffer, alternative routing does neither impact the client layer nor the server layer resources
substantially. Furthermore, the effectiveness of constrained alternative routing and of the shared
over ow capacity in OBTN is assessed by isolating them. This is achieved by comparing OBTNSummary iii
and BoCS which use the same virtual topology but differ in the routing exibility and network
dimensioning. The evaluations show that applying any of the concepts alone to BoCS does not
yield any or only limited improvements or even produces severe penalties. However, if applied
together as OBTN, they harmonize and effectively improve performance.
Evaluations with different FDL buffer architectures and dimensionings quantify the trade-off
of improved contention resolution, thus reduced node and network resources and the additional
node resources required for the FDL buffer. The results show that increasing the number of
FDL buffer ports up to approx. 16 24 leads to fewer node and network resources. Beyond this
dimensioning, node resources increase with diminishing reductions regarding resources in the
underlying server layer. Extending the studies to advanced virtual topologies, it is demonstrated
that OBTN is not restricted to the full-mesh virtual topology used for all previous evaluations.
A traf c demand-based and a path-length-based virtual topology design approach is introduced
and investigated. For one example parameterization, the number