Techno-economical analysis of some optical network architecture
Results from EURESCOM P615
(Detailed version)
T. Jakab1, D. Arató1, P. Fonseca2, J. Orfanos3, A. Ruiz Cantera4, M Luyten5, D. D. Marcenac6, A. Hamel7
1: Hungarian Telecommunications Co., 2: Portugal Telecom, 3: Hellenic Telecom, 4: Telefonica I+D, Spain, 5: KPN Research, Nederland,
6: British Telecommunications plc, 7: France Telecom Corresponding author: T. Jakab
MATÁV BME HT Sztoczek u 2. Budapest H-1111, Hungary
E-mail: jakab@hit.bme.hu Tel.: +36 1 463 10 10 Fax.: +36 1 463 32 66 Abstract
The main scope of the EURESCOM P615 project is to study the evolutionary path in the core and metropolitan area networks towards an optical network layer, describe, model and analyze optical network architectures in details.
Optical ring and mesh architectures are analyzed and compared in this paper.
Architectures to be studied were selected to have different combinations of the routing and protection functions realized in the electrical and/or optical domain.
Two standard SDH architectures (MSSP ring, DXC based mesh) were included in the analysis and the comparison, as references.
Several aspects like physical size, first investment cost, availability, OAM are covered in the analysis. A set of realistic network examples (with different topologies and demand patterns) were specified to support the analysis.
1 Introduction
The application of optical point to point transmission systems are wide-spread in today’s telecommunication networks. However, fundamental network functionalities (routing, protection) are realised mainly in the electrical domain (with SDH or SONET technology). Since there is a huge amount of fibres in the networks it is obvious that any network operator is highly interested in exploiting this potential. Wavelength division multiplexing (WDM) has been recognised as the key technology to upgrade existing fibre plants to real optical networks. This technological evolution offers additional functionalities due to the applications of optical nodes.
Numerous partly or fully optical network architectures based on different optical node configurations can be specified. EURESCOM P615 project titled Evolution towards an optical network layer is focused to study the evolutionary path in the core and metropolitan area networks towards an optical network layer, describe, model and analyse optical network architectures in details [1, 2, 4]. To identify most promising networking applications and the expected benefits of these optical architectures detailed techno- economical analysis and comparison of candidate optical and existing SDH architectures (for reference) were elaborated [3].
Based on the analysis results, background of the differences in the performance of the architectures are identified. Comparisons provide information on the best fit of the architectures to different networking applications and helps to select candidate architectures for the introduction and development of the optical network layer [4].
2 Studied architecture
In EURESCOM P615 project the evolution of the optical network layer is interpreted as the step by step introduction and implementation of the fundamental networking functionalities like routing and protection in the optical network layer.
Architectures representing solution of different combinations for electrical/optical routing and protection were selected for detailed modeling and analysis. A brief summary on the selected architectures is given in this section. Table 1 summarizes the main features of the studied architecture.
Table 1 Overview of studied architectures
Architecture Routing Protection Key equipments
Electrical Optical
SDH MSSP Ring electrical electrical multiplex section shared protection
ADM not present
CS Ring optical + elect.
electrical linear MSP improved with logical node reordering
ADM OADM OMSSP Ring optical +
electrical
optical multiplex section shared protection
ADM or DXC
OADM and optical switches SDH DXC Mesh electrical DXC based restoration (electrical) DXC not present MWTN Mesh optical +
electrical
OXC based restoration (optical) DXC OXC
2.1 Ring architectures
Two promising optical ring architectures and the standard SDH Multiplex Section Shared Protected (MSSP) ring - for reference - are studied in this paper. One optical ring architecture realizes routing and protection partially in the electrical and partially in the optical domain (Colored Section Ring) reutilising standard SDH equipment. Both the routing and the protection are realised in the optical domain for the other optical architecture (Optical Multiplex Section Shared Protected Ring).
2.1.1. Colored Section Ring
The basic idea of the Colored Section Ring (CS ring) architecture is to take advantage of the well known linear multiplex section protection (MSP) protocol available in standard SDH ADMs and the wavelength routing in a two-fibber bi-directional optical ring to increase the transmission capacity [5].
CS ring is installed on two fibres, the nodes are equipped with optical add-drop multiplexers (OADM) and standard SDH ADMs, however there are duplicated aggregate units (optical line terminations) in SDH ADMs due to the applied protection. Different wavelengths are assigned to each multiplex section interconnecting the SDH ADMs. Based on wavelength routing logical ring with node order different from the physical (cabling) one can be realised. Additional electrical routing to realise transmission demands between nodes not neighbouring on the logical layer are provided in SDH ADMs. Linear MSP is applied to protect the architecture against cable cuts: aggregate signals are splitted and routed via complementary arcs of the ring. Proper node order depending on the realised demand pattern can be established on the logical network layer to eliminate the need of electrical routing in intermediate nodes. In such a case node failures affect only originating and terminating traffic, thus the performance of linear MSP is significantly improved [6].
2.1.2. Optical Multiplex Section Shared Protected Ring
Optical Multiplex Section (OMS) Protection involves the simultaneous switching of wavelength multiplexed traffic from one fibre to another to avoid a broken fibre or a line
amplifier failure. In general it should be more cost effective than Optical Channel Protection where optical switches have to be provided for each individual connection.
The idea of the OMS protection ring architecture has been elaborated in frames of EURESCOM P615 project from the functional modeling to the physical implementation [4, 7]. Nodes in OMSSP ring are equipped either with SDH DXC or ADM besides the optical add-drop multiplexer. With an OMS protection ring architecture, traffic is switched at both ends of the broken Optical Multiplex Section from one fibre to the another counter- propagating fibre so that it goes around the ring to avoid the break. Note that dual-ended switching is required for any practical OMS Protection ring.
In the two-fibre OMSSP ring the capacity of each fibre is divided approximately equally between wavelengths which support working optical channels and wavelengths reserved for protection purposes. Protection wavelengths on the clockwise fibre provide protection capacity for the working optical channels on the counter-clockwise fibre and vice versa. If working optical channels do not occupy the same wavelengths on both fibres then optical protection switching can be achieved without resorting to wavelength conversion. Note that a two-way pair of connections between two nodes will use different wavelengths if they use different fibres.
This configuration means that in general any pair of nodes on the ring can be interconnected so that both connections can use the same route (but different fibres). This will normally be the shortest route in terms of number of en-route nodes, and hence this scheme frees capacity at the same wavelengths for other connections on the ring, i.e. the wavelengths may be reused for other connections. By comparison, in a two-fibre OMS Dedicated Protection ring working traffic is restricted to one fibre, so that each pair of connections between any two nodes requires an exclusive wavelength. Therefore the OMSSP ring generally offers a higher capacity than the OMSDP ring, all other conditions being equal.
2.2 Mesh architectures
Two mesh architectures, an optical and a standard SDH mesh architecture (DXC based mesh) are selected for detailed analysis. The Multi-Wavelength Transport Network (MWTN) project, part of the RACE program, set out to develop ideas for a future broad- band flexible transport network employing the optical network layers. The fundamental building block of the MWTN architecture is the Optical Cross-Connect (OXC). The MWTN OXC node has the ability to route traffic according to wavelength. Inbound traffic is selected first by its incoming route, and second by wavelength. Optical cross-connects are then used to redirect this traffic onto outgoing routes. Therefore each optical channel can be selected and redirected as required. In addition, optical channels may be added or dropped using a standard SDH cross-connect.
3 Aspects and the general approaches of the analysis
Several aspects like physical size, cost, availability, OAM are covered in the analysis. A set of realistic small network examples (with different topologies and demand patterns) were specified to support the analysis.
Typical services and network structures has been taken into account specifying the demand patterns. Typical patterns with normal and high values are specified to evaluate both the traffic capacities and the suitability to different traffic patterns of the different architecture.
For each network, three different traffic patterns, at two different traffic volumes, are specified. The three patterns are:
1. The "nearest neighbour” pattern is consistent with traditional telephony. The traffic demand drops with distance.
2. A hubbed traffic demand, consistent with the case where all the nodes send traffic to and from a hub node, which may be parented onto the higher tier of a network.
3. Evenly distributed (uniform) traffic, where each node communicates in equal amounts with every other.
To deduce trends of relative costs versus amount of traffic, the dimensioning should be repeated with 5 times the basic initial volume. Thus the normal values vary from 4 to 20 STM1s, the high values are 5 times higher.
In the specification of topologies (assumed existing fibre infrastructure) the typical sub-network (cluster) sizes and span length in metropolitan area and long distance networks were taken into account. Two topologies have been specified for the general study cases (rings and meshes), a 5 node topology with 10 km length for each span and a 8 node topology for with span lengths vary from 40 up to 100 km.
To perform the analysis first the different architectures are dimensioned to realise different network examples. With help of detailed dimensioning results statistics on needed network resources were elaborated. Based on this statistical informations the investment cost to realise different network examples can be calculated. Availability analysis is focused on the down time ratio (DTR) of the transmission routes in the different architecture. Two approaches were taken: a simplified one is dedicated to study single STM1 capacity transmission routes on the different architectures, and a more sophisticated one to analyse the availability performance of the dimensioned small network examples on the overall network level. Both analysis are based on the calculations of down time ratios of the available transmission paths of the architectures taking into account the implemented protection techniques.
4 Analysis results
For about one hundred small network study cases were analyzed and a large amount of detailed results were elaborated in frames of EURESCOM P615 project. In this paper only some selected results are presented to illustrate the achievements major conclusions are based on. For detailed results see [4].
4.1 Dimensioning
In network dimensioning the two optical rings (OMSSP ring and CS ring) are considered with STM-16 line systems, however, besides STM-16 MSSP rings, STM-64 MSSP ring is studied, as well.
Fibre length need of different ring architectures (in km) can be depicted on Chart 1 in a five node network example with different demand patterns. As it is expected due to WDM technique optical rings are with significant optical fibres savings compared with STM-16 MSSP ring. STM-64 MSSP ring is with similar fibre needs in most cases then optical ones, however this high speed SDH ring needs slightly less fibre in case of some specific demand patterns. (In the fibre savings calculation maximum 16 wavelengths per fibre is supposed for optical rings.)
Both optical rings provide significant fibre savings compared with STM-16 MSSP ring, in some cases depending on the traffic patterns OMSSP ring (equipped with SDH DXC) is much more powerful in routing then CS ring with SDH ADM. Thus, OMSSP is with slightly better performance from fibre savings point of view in these cases.
Fibre length savings (normal traffic, 5 nodes)
0 200 400 600 800 1000 1200
near.neigh. hubbed uniform
CSR OMSSP MSSP (16) MSSP (64)
Chart 1 Fibre lengths need for rings in a 5- node network example with normal traffic
Fibre savings performance of CS ring can be improved with help of the extra flexibility provided by the logical ring layer (node reordering with node duplications) [5].
4.2 Cost analysis
Based on the results of detailed network dimensioning first investment costs for different network examples are calculated.
In general electrical equipment cost savings are expected first of all in the intermediate nodes due to the optical routing, and optical restoration in meshed networks applying optical network architectures. Further saving can be achieved with help of the additional flexibility provided by the optical network layer, and the wavelength routing.
Architecture overall cost comparison (normal capacity, 5 nodes)
0 200 400 600 800 1000 1200 1400
near.neigh. hubbed uniform
Traffic pattern
MSSP (16) CSR OMSSP
Chart 2 Total equipment cost for rings with different traffic patterns in a 5-node network example
As it can be depicted on Chart 2 CS ring is less expensive then OMSSP, however OMSSP much more flexible for network growth (DXC, logical mesh).
Since network elements realized with passive optics (splitters, filters, etc.) are with low costs the contribution of electrical equipments dominates in the total equipment cost of the optical architectures. Since routing and protection partially or total realized in the optical domain the dominating electrical equipment are the termination equipment interfacing the network with the sources of the signals to be transmitted.
Only equipment costs are presented in this study, because it is not obvious how to calculate cost savings on less utilisation of already installed optical fibres (fibre savings).
Studying the cost breakdown of the total equipment installation costs (including fibre costs) the trade-off between fibre costs and optical equipment costs can be identified.
As it can be depicted on Chart 3 for high capacity demands comparing the SDH MSSP ring with optical rings the fibre costs significantly decreased, and the cost of needed optical equipment to install WDM systems are less then the fibre cost savings: thus it is clear that less expensive to install WDM systems then to install new fibres (infrastructure costs - like duct - are not included in the fibre cost).
The reason behind the relatively high electrical equipments for OMSSP is SDH DXC in the nodes, which equipment is more expensive then the SDH ADMs in CS ring. CS ring is with less electrical equipment costs, because of effective capacity utilization due to the minimized transit traffic in the intermediate nodes.
0 2000 4000 6000
Overall Cost Comparison (KECU) for a 5 node ring with uniform traffic
Fibres
Optical equipment Electrical equipment
5 times Normal traffic demand SDH CSR OMS-SP
Normal traffic demand
SDH CSR OMS-SP
Chart 3 Investment cost of three architectures for normal and high capacity uniform traffic.
Comparing the mesh architectures (Chart 4) MWTN less expensive then the SDH DXC mesh for study cases with high capacity demands (STM-16 line systems are supposed in both architecture).
The cost breakdown of MWTN (Chart 5) shows similar picture that for the optical rings. However, a relatively higher contribution of optical equipments can be identified because of the applied optical cross-connect switch is a more expensive network element then the optical add-drop multiplexers applied in optical ring architectures.
5 nodes SDH-MWTN total cost
0 2000 4000 6000 8000 10000 12000 14000
5 nn_n 5 nn_h 5 h_n 5 h_h 5 u_n 5 u_h
Total cost (kECU)
SDH MWTN
Chart 4 Total cost comparison of mesh architecture in a 5 nodes example
electrical / optical / infrastructure breakdown of MWTN architecture
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
5 nn_n 5 nn_h 5 h_n 5 h_h 5 u_n 5 u_h 8 nn_n 8 nn_h 8 h_n 8 h_h 8 u_n 8 u_h infr.
opt. int.
elect.eq.
Chart 5 Electrical / infrastructure / Optical equipment breakdown of installation costs for MWTN
4.3 Availability analysis
From the results of the availability analysis the study on two single STM1 demands routed on single and double span routes in 5 node network example are presented. The availability is evaluated calculating the down time ratio of the studied transmission routes.
The results are presented on Chart 6. The presented study cases are (from left to right):
MSSP ring, basic CS ring, CS ring with node reordering, basic OMSSP ring, OMSSP ring with 1+1 HW protected tributaries, OMSSP ring with 1+1 HW protected aggregates, OMSSP ring with 1+1 HW protected optical line terminations, and OMSSP ring with 1+1 HW protected aggregates, tributaries and line terminations. The two bars per cases represent single and double span routes, respectively.
Availability performance of basic CS ring is sensitive for the number of hops on the transmission path because of the non-protected intermediate node failures.
Additional flexibility based on wavelength routing provides good feature to improve standard linear MSP protection scheme implemented in the electrical domain.
Simple and effective optical protection against fibre cuts and line amplifier failures in OMSSP is realized in the optical domain on architectural level. However, there is no protection against the failures of the electrical network elements, thus any electyrical failure is critical for the basic OMSSP ring. Applying 1+1 HW protection on electrical equipment level to complete protection significantly improves the availability performance of the architecture [8].
DRT results of different study cases with an STM1 demand
0.0000E+00 1.0000E-05 2.0000E-05 3.0000E-05 4.0000E-05 5.0000E-05 6.0000E-05 7.0000E-05
MSSP CSR CSRREO OMSSP OMSSP t OMSSP a OMSSP lte OMSSP all
study cases
DTR 1span
2span
Chart 6 Availability of a single span and a double span 1 STM-1 capacity transmission routes on different optical ring architectures (5 node example)
Based on detailed studies electrical equipments are the dominating elements defining the availability performance of the optical architectures. For a single STM-1 demand MSSP and improved optical architectures with same availability performance, for higher capacity demands the optical rings are with better availability then the SDH ring.
4.4 Operation Administration and Maintenance
The main drawback of WDM based transmission network from OAM point of view that these type of networks need extra spare parts due to the different wavelengths applied in WDM technique.
Two detailed studies were elaborated in frames of EURESCOM P615 project from OAM point of view. One study was focused on the possible application of distributed restoration algorithms in optical mesh networks. Simulation has been performed to calculated the restoration times can be achieved in optical mesh networks. Restoration with and without wavelength conversions were analyzed.
Chart 7 shows three runs of the studied restoration protocol. In all three runs all the disrupted traffic is rerouted, although the time needed is different. The amount of spare capacity is determined in the initial planning of the working traffic demand in such a way that all traffic can be restored in case of a cable cut. The simulations have shown however that there are case for which the percentage is lower than 100 %. In that case the restoration algorithm is not able to find the available capacity in the network to re-route the traffic within a certain period of time. The choice of the restoration paths is a complex matter, with maybe only one possible solution but many different ways of starting the restoration.
Building extra initial spare capacity in the network could solve this problem.
Percentage of restored traffic
Restoration time for a single link failure
0.00 20.00 40.00 60.00 80.00 100.00
70.00 90.00 110.00 130.00 150.00 170.00
time (ms)
run#1 run#2 run#3
Chart 7 Performance of restoration for a single link failure for SDH mesh network.(Three runs of the protocol are shown.)
The restoration mechanisms for the SDH and MWTN networks differ in the following ways: For the SDH mesh, restoration is done by the digital cross connects at a STM-1 granularity. Normally spare timeslots in point-to-point links between two nodes are used to reroute interrupted traffic in the event of a failure. For the MWTN network, the re- routing of optical channels is performed by optical cross-connects, with an STM-16 granularity. Cases with and without wavelength conversion have been considered.
Globally, and for the various network scenarios considered, the ranges of restoration times shown in Table 2 were obtained. From results of Table 2 it can be concluded that there is no significant difference in the restoration speed between SDH and WDM mesh networks. The cases without wavelength conversion show that 100% restoration was not possible, and so the extra degrees of freedom provided by wavelength conversion are important in allowing a higher degree of restoration.
Table 2: Processing time for restoration times and percentage of restored traffic for SDH and optical (MWTN) meshed networks.
Electrical mesh (SDH cross-connects)
Optical mesh (MWTN) with wavelength-
conversion without wavelength conversion percent. rest.. time
[ms]
percent. rest.. time
[ms] percentage rest.. time [ms]
5 nodes 100 % 86-175 100 % 80-94 ms < 100 % 122-166 8 nodes 100 % 80-313 100 % 82-92 < 100 % 116-263
A study was focused on the extension of standard MSSP protocol for optical multiplex section and compare the switching performance of the optical and SDH implementation (Chart 8).
The SDH ring protection is found to be faster than for the OMS-SPRing, due mainly to the slower switching speed of opto-mechanical 2x2 switches compared to the SDH electronics, but in all cases the switching times are less than the 50 ms currently the standard for SDH networks.
Protection switching times for SDH MSSP and OMSSP
0 2 4 6 8 10 12 14
Node failure Bidirectional signal fail
Unidirectional signal fail
Unidirectional signal degrade Failure cases
Time [ms]
SDH MSSP OMSSP
Chart 8 MSSP switching times for SDH and optical implementations in case of different failure cases
5 Summary and conclusions
Based on the detailed results of the analysis work performance in frames of EURESCOM P615 project some promising optical architecture were identified. Optical ring architectures seem to be much more mature then optical meshes.
Reduction of first investment costs of optical architecture compared with SDH architectures can run up to 70%. This reduction of investment costs caused by fibre reduction and replacement of electrical equipment by low-cost optical equipment. Savings on investments costs of optical architectures are increasing with an increasing traffic demand.
The modeling of several performance parameters have shown, that there are no significant difference between optical and SDH architectures.
Architectures with protection realized in the optical domain may need extra hardware protection to improve the availability of the applied (not protected) electrical network elements.
6 References:
[1] M. O. van Deventer, J. Ramos, L. Blain, R. Grant, B. Skjoldstrup: Node functionalities and architectures for the optical network layer, results from EURESCOM P615, Proceedings of the European Conference on Networks and Optical Communications 1997, vol. Broadband Access Networks, pp. 34-41.
[2] T, Jakab, M. Schiess: Techno-economical analysis of optical network architectures and their introduction into today’s networks, Proceedings of the European Conference on Networks and Optical Communications 1997, vol. Broadband Access Networks, pp 42-46.
[3] M. Schiess, R. S. Grant, L. Cucala, J. van der Tol, Chr. Zimmer: Introduction Scenarios for Optical Network Architectures - Results from EURESCOM P615, accepted paper for European Conference on Networks and Optical Communications June 1998, Machester, UK
[4] EURESCOM Project P615 Evolution towards an optical network layer see EURESCOM Web Site:
www.eurescom.de
[5] L. Blain, A. Hamel, T. Jakab and A. Sutter: "Comparison of classical and WDM based ring Architecture" Proc. of NETWORKS'96, Vol 2 pp. 607-612 , November 1996, Sydney, Australia.
[6] T. Jakab, D. Arato, A. Hamel: Availability Modelling and Analysis of Optical Transmission Network Architecture, 6th Int. Conf. on Telecomm.Systems, Proceedings pp 372-387, March 5-8 1998, Nashville, TN, USA
[7] R. Grant: “Optical protection in a WDM ring: from functional model to implementation” accepted for DRCN'98, 17-20 May, 1998, Brugge, Belgium
[8] T. Jakab, D. Arató, D. D. Marcenac: “Availability Analysis of Some Optical Ring Network Architectures - Results from EURESCOM P615 Project accepted for DRCN'98, 17-20 May, 1998, Brugge, Belgium
