Converged optical network architectures

Transcription

1 Converged optical network architectures Ioannis Tomkos High Speed Networks and Optical Communications - NOC Research Group 1 Markopoulou Ave., PO. BOX 68, Peania, Athens, Greece

2 Outline Highly-demanding bandwidth-hungry and QoS-sensitive new services are demanding new networking solutions and network design paradigms What solution is in our hands currently? Network design: past & future Where things are heading to? Optical network architecture evolution What are the different areas where convergence might be required? Relevant research activities and results on topics related with convergence (AIT s activities) TRIUMPH, DICONET and ACCORDANCE EU research projects Summary 2

3 Optical technology - past 3 Optics up to 2000: Transmission plant capable of Tb/s transmission Hundreds of λs Hundred to thousands of km transmission in-between regenerators 2000 today Reconfigurable all-optical networks Advanced devices: Reconfigurable Optical Add/Drop Multiplexers (ROADMs) Wavelength Selective Switches (WSSs) Longer capacity x reach product at high spectral efficiencies Novel modulation formats and transponder concepts DPSK, DQPSK Coherent systems Optical technology finally reaches the access part Point-to-Point Ethernet Passive Optical Networks (PONs) What s next?

4 Future capacity demands New types of services emerging in Access High speed access via FTTx, DSL or wireless reach content web pages (e.g. flash enabled) large volume s (e.g. gmail) Voice over IP applications (e.g. Skype) Video distribution services (e.g. YouTube) Interactive multi-player gamming Video on demand (e.g. Triple-play services) Need for high capacity connections Access Metro Problems How to further optimize the network operation and reduce energy requirements? How to upgrade networks in order to meet future demands? How to achieve compatibility with existing solutions in order to: Ease control 4 Reduce upgradeability cost Even larger volumes of data carried in core networks

5 Network design - past In order to make our work easier, we split the network in different pieces: layers (physical, network, service, etc.) segments (access, metro, core) switching technologies (packets, circuits, etc.) administrative domains (multi-domain networks) technology domains (optical, wireless) Such divisions enabled optimized designs within each different piece of the network. However, in some cases the level of abstraction taken as a fact for the other pieces of the network by network designers was oversimplified and far from being realistic! In some cases there was no consideration what so ever on the issues that arise when you try to put together the different pieces of the network and what you have to pay in return for such suboptimal approach Lack of network efficiency (e.g. under-utilization of resources, stranded bandwidth, reduced QoS) Large power consumption Significant OPEX What s next? 5

6 Network design - future Removing the segmentations among the different pieces of the network Cross-layer optimization (physical/network, network/service, etc.) Multi-granularity switching nodes (OCS/OFS/OBS/OPS) Introducing end-to-end network solutions by removing the boarders among access, metro and core network segments Introducing end-to-end network solutions by allowing more flexible and efficient interconnection among different administrative network domains Introducing end-to-end networks solutions across technology domains (optical/wireless) Develop Converged Network Infrastructure ensuring further optimization and benefits for the users and services Lower CAPEX/OPEX Low power consumption Better QoS/QoE Support for new services Ease of network operation/administration 6

7 Future network architecture - EU vision 7

8 Evolution in Optical Network Architectures Network architectures can be differentiated in terms of o-e-o regenerators used administrative domains interworking switching scheme used network layers network segments used technology domains coexistent The aforementioned categories denote also the possible areas of convergence Some overlap among the different categories exist Can we find a solution that can be the convergence enabler across all different categories? 8

9 Network architectures in terms of o-e-o regenerators used 9 Opaque (o-e-o everywhere) Transparent (o-e-o nowhere)

10 Network architecture approaches in terms of regeneration needed others Infinera 10 R. Wagner, LEOS 2000, TuC1, November 2000 Future high capacity networks require an efficient, flexible and intelligent optical transport layer?

11 Limitations of transparent optical networks In transparent networks signal impairments accumulate : Dispersion ASE noise Fiber nonlinearities Crosstalk PMD Appropriate pre-planning may improve the window of operability In dynamically reconfigurable networks the window of operability is significantly reduced: The temporal variability of the transmission parameters prohibits efficient static pre-compensation of the impairments Failures propagate in a transparent network environment and therefore they can t t be easily localized and isolated 11 Advanced performance monitoring techniques will have to be implemented

12 Overcoming the problems Use of optoelectronic regenerators on per channel basis enabling impairment compensation and BER monitoring Use of dynamic impairment management techniques that may be implemented in-line (e.g. optical means of impairment compensation) or at the optical transponder interfaces (e.g. electronic mitigation of impairments) Some techniques require performance/impairment monitoring (optical or electrical) for offering dynamic operation In addition to physical layer impairment management techniques, the network designer may use certain Routing and Wavelength Assignment (RWA) algorithms that take into account the signal impairments (IA- RWA) and constrain the routing of wavelength channels according to the physical characteristics of the optical network paths. Such capability can be used as a tool for improved efficient network design and planning of advanced mesh optical networks 12 See tutorial by S. Sygletos, I. Tomkos, J. Leuthold, at J. Opt. Networking 2009

13 Transparent Optical Networks The introduction of more optical transparency in the network promises a series of advancements in the way the network operates and the benefits expected by end-users and services However realizing an end-to-end transparent optical network is very challenging and requires several new technology innovations (once realized though, it should be very easy to manage) Novel modulation formats (OFDM?) and amplification schemes All-optical (multi-wavelength!) regeneration All-optical sub-wavelength traffic switching/grooming New algorithms and control plane extensions 13 Alternatively a translucent (managed-reach) semi-transparent network may be an alternative option, which however requires a lot of complexity in managing the infrastructure

14 Network architectures in terms of administrative domains interworking 14 Single domain (isolated due to different operators, ) Multiple domains (interworking)

15 Multi-domain networking Multi-domain aspects in transparent/translucent networks have not been considered in detail so far Current developments have been mostly conceived for single-area networks While there will be a must in the near future, the protocols/techniques for achieving truly end-to-end optical connections (spanning different domains) have not yet been devised. Efforts are being initiated or are underway on the development of new features enhancing the GMPLS control plane performance. An integrated GMPLS based multi-layer and multi-domain control plane, which allows a more efficient use of the network resources, is a key research objective for coming years. 15

16 Control plane elements to address multi-domain issues 16

17 Multi-domain issues To achieve end-to-end lightpaths establishment in multi-domain networks, several domains should cooperate at the data, control and management plane levels and following a well defined yet diverse number of interconnection and information exchange models. Some key issues with respect to inter-domain networking are: How to exchange networking information between domains for routing purposes. How to negotiate QoS parameters between domains taking into account the heterogeneous mix of services and the particular policies applied. How to ensure coordinated resilient operation. How to provide interfacing between different technologies. How to interface control planes, which are usually running on a single domain The different interconnection models, defined at specific reference points such as the User-Network Interface (NNI) or the External and Internal Network-Network Interfaces (E-NNI and I-NNI), are characterized by the type and amount of information that is exchanged at network domain edges, by the applicable control procedures and by the service selection and activation methods. 17

18 Network architectures in terms of switching scheme used Circuit switching Packet switching 18

19 Convergence of switching techniques The network evolution mandates a framework that can provide control and management of traffic (dynamic bandwidth allocation) in multiple levels of granularity (fibre / waveband / wavelength / burst / packet / timeslot) A convergence of the various existing optical switching techniques (optical circuit/flow/burst/packet switching) is mandatory in order to achieve optimal exploitation of the network resources. Hybrid switch implementation (combining all switching schemes) A switching scheme that can adapt to any traffic characteristics/requirements (Polymorphous OBS?, OFS?) Observation: All techniques that achieve fine sub-wavelength granularity utilize the time dimension to allocate bandwidth to different users/services. Can the frequency domain being used instead? 19

20 Optical circuit switching for core networks In the backbone networks, clearly optical circuit switching is so far the winner Can it continue to be in the future? How it should adapt? Wavelength routed all-optical switched networks is the current state of the art under the optical circuit switching paradigm Now: Lightpaths Future: Lightrails, lightrees, etc. However Optical Time Division Multiplexed networks might be a future solution Enabler for 1TbE? 20 How to enable trans-multiplexing and interoperability?

21 TDMA based access networks In the access part of the network TDMA based protocols are dominating E-PON G-PON Can they continue to scale beyond 10G symmetric? WDM-PON is trying to make a paradigm shift from time-domain to frequency (wavelength) domain OFDMA-PON proposed by NEC is also using the frequency domain in somewhat different form 21

22 Network architectures in terms of network layers 22 Single layer optimization (physical, network, service) Cross layer optimization (physical/network, services/network)

23 Network layers convergence In order to handle the huge available capacity in the most efficient and dynamic way a convergence between multiple layers in the protocol stack - namely physical, network, transport and application - is imperative. 23 This involves designing the optical network with the main concern being the quality of the services that will be carried over it and includes optimization of protocols belonging to multiple levels of the stack for their best possible interoperability e.g. providing services requirements to the network design (service oriented optical networking) e.g. adjusting TCP functionality or parameters to enhance the operation of the underlying switching schemes e.g. designing the optical switching protocols in a way that respects specific service performance requirements and reconciling concepts such as IP over optical networks e.g. optimizing the network design by considering the physical layer capabilities of the network (impairment-aware optical networking more details follow in the next slides)

24 Impairment Aware Optical Networking Receive performance monitoring information related with impairments in the network or/and calculate the impact of impairments Run routing and wavelength assignment algorithms that take impairment information into consideration (IA-RWA) + Impairment aware networking 24

25 IA-RWA in transparent networks IA-RWA can be used either for pre-planning planning purposes (simple) or for dynamic network optimization (demanding) An operator who wants to avoid having to provide impairment- related parameters to the control plane may elect to treat them at the system design and static pre-planning planning level. In this approach the operator can pre-qualify all or a set of feasible end-to to-end optical paths In dynamic network optimization the calculations or/and measurements should be performed on real time and make use of the current state information aiming to minimize the blocking probability of new connection requests while guaranteeing the performance of the established connections 25

26 Challenges for the development of cross-layer optimized network operation/planning How to collect the impairment information from the network? Requires the development of optical monitors How to model fast and accurately the physical layer impairments Speed is required in the on-line case Accuracy is required in the off-line case What is the optimal route? Requires development of appropriate algorithms able to calculate fast and accurately the optimal route How to inform the network? How to inform the network? Requires the investigation of control protocol extentions (based on extending current protocols options, like GMPLS OSPF or/and RSVP) 26

27 The DICONET consortium DICONET Dynamic Impairment Constraint Optical Networking 27

28 DICONET project scope and solution framework DICONET scope The development of a dynamic network planning/operation tool residing in the core network nodes that incorporates real-time measurements/calculations of optical layer performance into IA- RWA algorithms and is integrated into an IA- control plane DICONET key elements Physical layer monitoring Physical layer modeling Physical layer aware RWA Physical layer aware GMPLS 28

29 Monitoring: OIM & OPM Techniques WDM Layer Parameters Aggregate power Channel Power Channel wavelength Spectral OSNR Signal Quality Parameters In-band OSNR Q-Factor/BER Bit Rate Jitter D. Kilper et. al., Optical Performance Monitoring, JLT 2004 C. Mas, I. Tomkos Failure management in optical networks, (invited) ICTON, 2003 Digital techniques use high-speed logic to process digital information encoded on the optical waveform Cannot isolate the effects of individual impairments BER computation is not fast (complex, expensive) Analog techniques treat the optical signal as an analog waveform and measure its specific s characteristics Time domain : Eye diagrams Histograms Auto/cross correlation meas. Frequency domain : Optical Spectrum RF amplitude Spectrum 29

30 Signal Performance Modelling based on Q-Factor Q (Q-tool) Probability V (1) σ(1) Threshold voltage V th LOG (BER) BER = Q 1 Q e 2 erfc 2 2 Q 2 2π -10 σ(0) V (0) Threshold Voltage (mv) Random amplitude perturbations are considered with Gaussian distributions Q end I I I I σ + σ 1 end 0 end 1 end 0 end 1, start 0, start = = σ1, end σ 0, end I1 I start 0 σ start 1, end σ + + 0, end = ( Eyeimpairments) ( Noiseimpairments) Q start Q start 30 Extensive literature on analytical Q-factor modeling considering all impairments (NRZ & RZ OOK)

31 Reference Network National topology based on the Deutsche Telecom (DT-net) Network Parameter Number of Nodes Number of links Value Node degree 3.29 (min: 2, max: 6) Link Length (km) 186 km (min: 37, max: 353) 31

32 Off-line planning performance results Real Traffic DTnet - Real traffic load Running times for W=50 Formulation running time (sec) RWA 476 RWA-p Sigma Bound 4610 IA-RWA 2 manages to pass all connections with W=50 and also have acceptable running time 32

33 Online network planning issues Once connections are established and the network is operational, how to decide how to handle a new connection request? The new lightpath that will be established, will affect the other already established lightpaths Thus, when a lightpath is established the quality of transmission of some already established lightpaths may become unacceptable These connections have to be rerouted! Rerouting a connection is a process that we want to avoid it involves tearing down the previous lightpath, re-executing the algorithm and establishing a new lightpath all these actions would interrupt the service of the connection and possible affect the quality of service exhibited by the end users 33

34 On-line planning performance results DT network Poisson arrivals (λ=1), exponential duration (1/μ=100) Three algorithms developed and evaluated according to: Blocking probability Average number of reroutings per request Acceptable execution times (~msec per request) Blocking probability Rerouting probability MUW bq-muw 0.05 bq available wavelengths MUW bq-muw bq MUW: Most Used Wavelength bq: Better Q Performance bq-muw: Mixed better Q and wavelength utilization (bq-muw) available wavelengths 34

35 Towards IA Control Plane There are generally three approaches to incorporate physical layer impairments information in the control plane: Routing protocol based Signaling protocol based PCE based What are the changes that need to be made in order to turn those approaches to an Impairment Aware Control Plane? 35

36 IA Control Plane - Signaling approach Extensions needed to RSVP-TE protocol 36

37 IA Control Plane - Routing approach Extensions needed to OSPF-TE protocol 37

38 PCE (Path Computation Element) approach 38

39 Network architectures in terms of network segments 39 Multiple segments (Access, Metro, Core) Single segment

40 Network Segments and characteristics Long Haul: National or regional intercity longer span Distances: 100 s to 1000 s of km Rates: from 2.5 Gb/s to 40 Gb/s 40 to 80 DWDM Channels POP Access Feeder POP Regional: City to city (little grooming) Metro: Office-to-office, switch-toswitch traffic within city (grooming) Distances: 10 s to 100 s of km (up to 400) Rates: from 633 Mb/s to 2.5 Gb/s ATM / Sonet, GbE, up to 40 DWDM Ch. Access: Residential access (CO or Head-end to residential subscriber) Distances: 10 s or less Rates: from OC-3 (155) to OC-12 (633) Sonet & ATM, GbE, CWDM? Premises: Customer LAN 40

41 Network segmentation trends Currently the different network segments (access, metro, core) have different traffic characteristics to serve, different cost points to satisfy, different bit-rates and transparent reach to handle However, it appears that the metro segment most likely will vanish in the near future Core networks will be connected directly to extended reach access networks, thus minimizing the number of central offices and cabinets/base-stations that the operators have to manage (resulting in less power consumption and OPEX costs) Present optical access systems mainly use PtP and PON technologies with a trend to extend reach and split which enables reduction of POPs in the access area. Thus, future access nodes will be required to handle multiple Gbps of traffic, making the concept of a converged (even from a topological point of view) access/metro/core network seem logical, if not unavoidable. 41

42 But what will be needed to transparently cross network segments? 42

43 Lessons learned from SDH/SONET How grooming works Grooming requires a mix of space switching (port-to-port) and time domain switching (time slot to time slot). This allows for the most efficient utilization of network bandwidth. The optical switching systems need to employ grooming capabilities to ease market entry. 43

44 The TRIUMPH consortium TRIUMPH Transparent ring interconnection using multi-wavelength photonic switches 44

45 TRIUMPH Project Concept High capacity network with trans-parent connection between metro / core rings (130 Gbit/s per λ-channel) and metro / access rings (43 Gbit/s per λ-channel) Interconnection between rings requires edge node (TRIUMPH) Node functionality: Transparent traffic grooming with time slot interchange 2R multi-wavelength regeneration 45

46 Experimental Results 46 Results from lab experiments and field trials were presented in the post-deadline paper sessions of ECOC 2008 and OFC 2009

47 More Switching Scenarios Time slot interchange 47 Obtained similar performance for many more scenarios Potentially energy efficient TSI scheme (no buffers required!)

48 Network architectures in terms of technology domains coexistent Wireless Technologies Wireline Technologies 48

49 Wireline - wireless convergence The access part of the network has traditionally been the bottleneck of the whole system. DSL technology is reaching its limits Wireless access becomes more and more pervasive Fibre to the Home (FTTH) solutions are becoming more and more prominent Hence, considerable enhancements in the access part of the network are needed and include the convergence of wireline optical and wireless technology reaching a hybrid optical/copper/wireless access infrastructure that will facilitate user mobility and support the vast number of devices and sensors that will need to connect to the internet from the user premises 49

50 The ACCORDANCE consortium ACCORDANCE A Converged Copper-Optical Optical-Radio OFDMA-based Access Network with high Capacity and flexibility 50

51 ACCORDANCE research initiative ACCORDANCE (A Converged Copper-Optical-Radio OFDMA-based Access Network with high Capacity and flexibility) introduces a novel ultra high capacity (even reaching the 100Gbpsregime) extended reach optical access network architecture based on OFDMA (Orthogonal Frequency Division Multiple Access) technology/protocols (based on concepts introduced by NEC), implemented through the proper mix of state-of-the-art photonics and electronics. Such architecture is not only intended to offer improved performance compared to evolving TDMA-PON solutions but also inherently provide the opportunity for convergence between optical, radio and copper-based access (common PHY). ACCORDANCE hence aims to realize the concept of introducing OFDMAbased technology and protocols (Physical and Medium Access Control layer) to provide a variety of desirable characteristics, such as increased aggregate bandwidth and scalability, enhanced resource allocation flexibility, longer reach, lower equipment cost/complexity and lower power consumption, while also supporting multi-wavelength operation. 51

52 State of the Art on OFDM technology OFDM: Modulation method for better transmission properties of bit streams Utilizes several low bit rate sub-carriers of the link to carry different QAM symbols simultaneously OFDMA: Application of OFDM as a scheme allowing for multiple access i.e. different users assigned to different sub-carriers OFDM technology currently used in: Copper, in the xdsl links using DMT (Discrete Multi-Tone) modulation format Radio (WiFi:802.11a, g, WiMax:802.16e-2005, 3GPP LTE) Indoor Power Line Communications (PLC), with the HomePlugAV specifications Recently OFDM(A) makes its way into the optics world A few recent studies show that OFDM can provide high capacity, long-reach and cost-effective operation for Passive Optical Networks (OFDMA-PONs by NEC) Convergence of optical infrastructure with wireless solutions also employing OFDM technology seems a beneficial direction to go 52

53 Concepts of ACCORDANCE research initiative (1/2) Proposed architecture 53

54 Concepts of ACCORDANCE research initiative (2/2) Example of (a) downstream FDM window assignment under an Optical FDM/OFDM assumption and (b) assignment of individual sub-carriers to different ONUs at a given point in time. 54

55 ACCORDANCE Benefits Improvement of transmission properties Dynamic Bandwidth Allocation Virtualization of Resources Cost-effectiveness Novel business and tariff models Smooth migration from current technologies Wireline-wireless convergence 55

56 Summary Next generation networks are facing new challenges and they need to transform We discussed the future network design principles with a particular view on optical network architecture evolution The optical network architectures were categorized in terms of o-e-o regenerators used administrative domains interworking switching scheme used network layers network segments used technology domains coexistent In each category, the different areas where convergence might be introduced to benefit the future traffic characteristics/requirements were analyzed New research directions were highlighted Relevant research activities and results on topics related with convergence where presented 56

57 Thank you! 57 Acknowledgements for contributions in this presentation: Acknowledgements for contributions in this presentation: - All partners of TRIUMPH, DICONET, ACCORDANCE - Dr. D. Klonidis, Dr. Y. Pointurier, Dr. K. Kanonakis - Ph.D. Students: S. Azodolmolky, M. Angelou, M. Spyropoulou

58 For more information on Convergence IEEE Network Magazine Special Issue on Protocols and Algorithms for Future Cross-Layer and Hybrid Optical Networks (May 2009) IEEE/OSA JLT Special Issue on Converged Optical Network Infrastructures in Support of Future Internet and Grid Services (July 2009) Symposium on Cross-layer Optical Network Design at ECOC 2009 Workshop on Dynamic Converged Optical Networks at OFC