Bulky IP network systems have been in the recent times fast applied directly in favor over point-to-point DWDM systems. Nonetheless, with the rising use of Opaque Reconfigurable Optical Networks (RON), interest has been generated in the use of a RON in provision of the desired connectivity or links among IP routers. An opaque reconfigurable optical network is defined as a system of optical cross-connects (OXCs) which contains optical or electrical fabrics bounded by optical transponders enhancing opto-electronic regeneration coupled with a system of wavelength conversion. A RON avails fast provisioning applicable in responding to topical changes and expansion in IP traffic needs alongside speedy, cost effective re-establishment of failures in IP linkage. Consequently, a lot of research has gone into the exploration of the multiple facets of IP over RON design, such as experimental analysis, routing algorithms, re-establishment schemes, network control and its management, control protocols,signal generation protocols which ensure establishment of a dynamic connection linking IP routers over the RON (Zhang, 2003). Nevertheless, huge applications in favor of IP rather than reconfiguable optical network architectures have not yet been extensively implemented.
Huge ISPs characteristically apply a hierarchical IP network design model: single or dual huge backbone routers (BRs), positioned in key Points-of-Presence (PoPs), aggregate traffic starting from access routers which then forwards them to other access points. Access routers (ARs) are located together with backbone routers that support traffic starting and ending in that point; or remotely placed whereby they aggregate traffic originating and terminating in smaller cities. Traffic sourced from remote access routers can be connected to protected facilities though it is most commonly dual-homed in diverse routed unsecured SONET/SDH-over-DWDM circuits. Using dual routers in key PoPs in addition to applying intricately routed dual-homed AR-BR connectivity links ensure IP routers to re-establish traffic when faced by multiple linkage failures such as router problems, fiber cuts and failures associated with the router line card. This vital architecture is referrred to as Dual-Router Architecture as shown in Fig. 1. If every AR is however connected to only one single BR, as exemplified by a Single Router Architecture, once failures occur in the BR,isolation from the network of all traffic terminating in the associated PoP occurs. Therefore, such an architecture necessitates a ultra-reliable router (Zhang, 2003). This paper focuses more on the dual-router architecture given that realibility in routers for larger service providers remains a challenge.
Preliminary feasibility research into the potential financial viability of IP in relation to an optical network have previously been presented as illustrated by the the literature as: Reingold et.al (2002) analysed the single and dual router architectures and found that an IP architecture in comparison to a RON architecture is cost efficient in the single router architecture if only the system does not handle BR failures. Sengupta (2003) and Labourdette (2003) however contrast the findings that IP in comparison to RON designs are cost effective for both single and dual routers when compared under diverse IP virtual topologies for an IP over RON plus the IP over DWDM designs. In routing IP links directly over DWDM architectures, an assumption made by the authors is that the IP links which pass through an office terminate within the router, while in the IP over RON architecture, optical cross-connects refferd to as “express” links in offices are applied to enable IP links so as to circumvent the router to lessen the number of router line cards that would have otherwise been necessary. First, this comparison may be biased rather than justified since the “express” links facilitated by the optical cross-connect are currently being directly applied over point-to-point DWDM. Secondly, the assumption implied by the authors in both texts is that both RON topology and IP topology are identical which is however negated by the fact that the RON topology and the IP baseline topology are different with the RON network being commonly larger in the number of nodes and links required. Thus, reliability in a RON network can be ensured by building access links from an access router to the corresponding backbone router. Moreover, both Sengupta (2003) and Labourdette (2003) assumed that costs associated with various optical layer devices were negligible in their economic feasibility analysis in such as DWDMs, optical transponders, and optical-based amplifiers. In , two integrated IP/optical designs have been proposed which use additional router line cards which are connected to a RON whereby cost effective failure recovery is ensured by the tight integration amid the two layers. This have been denoted as the Hybrid and Modified Protected architectures which when contrasted with to the Baseline architecture, that is the direct connection in IP over point-to-point DWDM architectures, and the Protected architecture whereby every IP link is routed over the RON which enables optical network re-establishment; the preliminary findings showed that these incorporated IP/optical designs are the most attractive alternatives to such other architecture as the baseline architecture. Further modelling and a comprehensive analysis is therefore vital in exploring integrated IP/optical architectures.
Hence the feasibility findings in Reingold et.al (2002) shall be extended so as to build the results defined in the architectures posted in Li (2004). Further, four architectures which are Baseline, Protected, Hybrid, and Modified Protected have been furtrher analyzed and expounded more as compared to Li (2004). First, the initial implementation cost associaied with each of architecture has been evaluated by dimensioning the IP and optical networks for definite traffic matrices. Secondly, costing the requisite components is carried out which is then used to simulate the impact of traffic on the corresponding capital expense of each architecture. The networks are design with adequate capability to ascertain recovery caused by any event such as fiberspan and router malfunctions.
A. IP over Optical Network Architectures
In this case, the optical layer entails a DWDM transmission layer, fiber media layer and an optical switching layer. The DWDM layer consists of either point-to-point designs or ultra-long haul systems containing optical add-drop multiplexers. In such a layer decompositionthe design of an IP network over an optical layer is presented over three alternatives: one is routing IP connectivity links through hard-wired point-to-point DWDM systems then ensure restoration is enabled by the IP layer; or routing IP connections through an optical switch layer then ensure restoration through both layers; or finally, routing IP conections via Ultra-Long Haul (ULH) DWDM topologies which incorporate restoration capabilities which are enabled by IP and transmission layers. Only three layers are needed in the first and third alternatives while the second alternative necesssitaes four layers. The first two alternatives are more utised hence shall form a basis for this study. The four IP over optical network architectures can be expounde in details as follows:
1) Baseline Architecture
This corresponds to the first alternative as stated above whereby each IP link is hard-wired or permanantly routed among one or more point-to-point DWDM architectures. Therefore, the baseline architecture avoids all forms of optical and electrical cross-connect basis hence restoration within the optical layer is not provided.
Responsibility for all failure recovery is assigned to the the IP layer which applies routing protocols such as OSPF and IS-IS which “reconverge” when failure occurs to reroute traffic through possible alternative paths. IP routing protocols are known to be slow in failure recovery and take from ten seconds to minutes so as to recover from a single-event failure which is too long for delay-sensitive applications such as voice calls through IP or real-time Internet softwares. However, there is major continuing study which is concentrating on ways to reduce routing protocol convergence periods so as to speeden up recovery from failure (Riecke, 2001). Furthermore,means like MPLS fast re-route can be incorporated to enable faster failure recovery though this comes along with additional complexity (Pan, 2004)
Although the baseline architecture is simple,it brings about major capacity building and network maintenance difficulties in addition to coordination problems between the IP layer and the transport layer. Since optical layer links are insecure, the IP planning method must incorporate ample recovery capacity in the IP layer so as to overcome failures in the IP and transport layer. specificaly, IP networkdesigners must posses knowledge as to the physical layer routing design for every unprotected IP link which ensures links routed over various basic fiberspans can be identified in order to recover from that instance of fiberspan failure thereby complicating the planning process. Additionally, since the transport layer projected maintenance activities influence the IP traffic routing, its is crucial to co-ordinate a maintenance plan between IP and transport so as to ensure that an ample capacity to carry traffic is availed (ensures that a severe degradation in network performance does not occcur through the simultaneous maintenance on both fiberspans and routers) and that maintenance operations on fiber facilities does not necesitate the continous re-routing of traffic when fibres are pulled in and iut by maintenance workers (i.e., evading link flapping from occuring in the routing protocols). A carefully designed procedure and co-ordination plan should therefore be maintained between diverse work centres in a large ISP.
2) Protected Architecture
The Protected architecture corresponds discussed here-in corresponds to the second alternative that RON is found amid the IP layer links and the DWDM level. The RON enables restoration from failures in the optical-layer such as in the optical components and fiberspan cuts for the connectivity links for vital backbone routers. Generally,failure recovery times from linkage failures are reduced by the faster restoration through RON as compared to IP routing protocols. nevertheless, the RON cannot re-establish restoration for failures in the IP layer like router failures which can only be restored within the IP layer. Therefore, over and above the spare capacity availed in the RON so as to recover from IP link failures, more spare capacity is necessitated in the IP layer in order to enable recovery from router-related failures. Hence, there is a duplicate of spare capacity in this architecture in IP and optical layers.
Not only do protected IP links shorten the duration needed for recovery but also minimizes the operational complexities necessitated by various planned maintenance events and scheduling restoration network capacity as already illustrated. The protected architecture differentiates planning and maintenance events in the IP and optical layerswhich eliminates the need for critical coordination implementation between the two layers. This ensures various procedures are simplified such as maintenance operations between the IP and transport work centres which eliminates all necessity for any additional co-ordination processes and tools.
3) Hybrid Architecture
In the hybrid architecture, rather the approach whereby all IP links connecting BRs are carried over the RON, it applies the RON chiefly in the provision of an extra link capacity in the restoration process. similarly to the baseline architecture, IP layer links are directly conveyed over unprotected point-to-point DWDM designs. However,in contrast to the Baseline designs, the IP layer is exactly dimensioned with sufficient capacity for only carrying service traffic such that no spare capacity is availed for failure recovery in IP layer. There is therefore a need as illustrated in Pongpaibool (2006) to install spare router line cards in every BR which are linked to a co-located OXC found in the RON. In regular operation, the router spare line cards are left idle such that zero connections are made between IP routers over the RON. In the event of failure, the spare router connecting to the OXC links establishes a connection to the RON in order to increase the IP network capacity referred to as the ‘bandwidth on demand.’ This shifts restoration from the IP layer to the optical layer while further providing an opportunity for broader sharing for the current RON restoration capacity among IP layer functions and other services like Private Line services. In essence, router and fiberspan failures are commonly non-simultaneous occurences hence enabling sharing of the restoration capacity. Therefore, the restoration capacity held in reserve for fiberspan failures can be applied for cases of IP router failure.
Two network clouds exist: the upper cloud consists of a reconfigurable optical network which has a shared mesh restoration while the lower cloud entails several pure point-to-point DWDM applications. IP links in the customary service bandwidth are normally hardwired throughout the DWDM systems without factoring in restoration. Nevertheless, each router shall be installed while incorporating spare router line cards connecting to the OXCs. An assumption is hereby made that the optical network further provides transport services to clients having shared mesh restoration (Li, 2002). In all likelihood, the DWDM layer and the RON shall share a common physical fiberspan topology. Hence, a single failure in the fiberspan shall affect services not only in the point-to-point DWDM systems but also in the RON. This necessitates an additional capacity in the RON so as to handle all restoration necessary forfailures in the IP layer links. However, a single failure in the backbone router or router line card impacts only on the IP network hence the restoration resources applied in the RON for failures in the links in either IP and any other services could also be applied in restoring IP traffic resulting from such a failure. This implies that there must be adequate additional capacity in the optical network so as to tackle any failures in the IP network fiberspan, while failures in theIP layer can in actual fact be recovered cost-effectively. more spare router line cards are needed in this architecture which increases the total cost. However, elimination of the IP layer restoration with the reduction of the necessary router line cards and IP link capcacity is made possible thus decreasing the total cost.
Recovery from failures in both IP and optical in this architecture necessitates IP routing protocol reconvergence, and then a new optical layer connection should be established so as to provide the obligatory restoration capacity. New IP links could be rapidly established among routers if express signaling is applied in the RON design and link layer protocols are tuned appropriately (Li P. S., 2003).This scheme however requires IP routing convergenceor else it shall be evidently slower than the recovery times attained in the baseline architecture implying failure recovery times could be critical issues in the hybrid architecture. Further, this architecture has inherent complexities associated with the baseline architecture across layers in the management, planning and maintenance.
4) Modified Protected Architecture
The Modified Protected architecture attempts to adress the relatively slow recovery times associated with the hybrid architecture. Its design is similar to the hybrid architcture.However, IP service links are routed over the RON rather than direct routing over the unprotected DWDM level. the IP layer is strategically dimensiod with a sufficient capacity capable of handling all service traffic whereby all the links are transported over the RON. Optical failures such as fiberspan cuts are managed through the optical layer restoration thus speeding up recovery in comparison to the hybrid and baseline architectures. nonetheless, like the hybrid architecture, failures in the IP layer are recovered through spare router line cards which are linked to the optical network and a dynamic connection made over the RON. Therefore, this approach shifts IP layer restoration capacity onto the optical layer. Incase of a failure in the router, the routing protocols in the IP layer re-route the traffic, then the router automatically and dynamically requests fresh connections via the spare router line cards in conjuction with the RON so as to launch fresh IP links among spare router line cards which provides a restoration capacity. This approach is relatively slow but as routers become more reliable over time, router failures are projected to lessen while the restoration time also becomes less critical. failures in the Router line card are recoverable using swift 1:n local router line card breakdown recovery, whereby a single extra router line card is applied in recovering from the failure in any one of the multiple operational router line cards (Li P. S., 2003).
B. Network Dimensioning and Modeling assumptions
In the evaluation of the capital needed for implemention of each architecture,first, each network is dimensioned and then costing the requisite components. First, a general, pre-defined, fiber topology with the associated DWDM systems, through which IP and RONs are transported. Th assumption made is that the network topologies for IP and RON, the varying traffic requirements for IP and private line services, and the way each IP and RON link has been routed over tthe fiberspan topology are pre-defined and calculable. It is threfore possible to model, design and dimension the network in various diverse fiberspan failure scenarios.
To establish the necessary capacity needed in the IP layer and RON, first route the traffic in the network in order to dimension the demands of the service. The assumption made is that routing in IP has its base on Open Shortest Path First (OSPF) with links in the optical network routed while maintaining the fewest number of hops. the set of failure outcomes can be said to be satisfactorily considered such that each failure scenario can be enumerated in turn enabling the traffic to be routed in reaction to the failure. The dimensioned capacity via link channels amid each pair of adjoining routers and on every RON channel is at the utmost necessary level across all single failure outcomes. even if numerous simultaneous failures can happen in networks, they are normally very rare. The assumption in common practice in network design as applied here-in that is that there is a need to dimension adequate capacity to endure only a single failure which can either be a single router or fibrespan failure. The analysis here-in shall focus only on the backbone network which entails backbone routers and associated links. ARs and remote RARs total to a single virtual AR such that every PoP is modelled by three routers which consists of dual BRs and a single AR. The dual BRs are normally dual-homed to the single AR then further connected to other BRs through backbone channels.
In this study, definite details on network dimensioning applied in diverse architectures are as follows:
1) Baseline Architecture
In dimensioning the baseline network, the IP topology is fixed ( this entails router placements, allowable BR-to-BR links, and routing links via point-to-point DWDM designs) with AR-to-AR traffic matrix coupled with OSPF weights. Under the standard IP network design methodology, every IP link comprises of two maximum traffic utilization constraints: non-failure state situation in the network and the failure state. The highest acceptable link utilization in the occurrence of failure is normally equal or to some extent higher than the maximum tolerable utilization during non-failure occurences. A higher utmost satisfactory link utilization throughout failure is often seen as an suitable concession between the additionalexpense of the spare restoration capacity and performance when the relatively rare failures occur. The traffic is then routed and computed under the utilization constraints whereby the ensuing capacity vital between every pair of adjacent routers is designed to cover the utilisation constraints. this is repeated for every failure possibility, such as all router and failures in the fiberspan, dimensioning every link to the maximum capacity mandatory over all failure occurences. Afterwards, the network cost is determined by addition of all contributions emanating from optical transport systems and router line cards which entail optical transponders, optical amplifiers and DWDM terminals. The assumption made is that the expense incurred in router chassis is amortized into the costs accrued by router line cards.
2) Protected Architecture
In modeling the Protected architecture, a similar IP topology, OSPF weights and traffic matrix are applied in dimensioning this architecture. A sufficient capacity is then dimensioned at the IP layer capacity in order to enable recovery from every instance of single router failure. The IP links are then routed over the RON, and therefore computations of the RON restoration requirements are possible (which are used in recovery from any fiberspan failure). Under this architecture, it is noteworthy that each IP link may transmit multiple IP OC48 or OC192 channels.
An assumption that can be further be made is that optical network applies shared mesh restoration whereby every optical connection entails two paths: a vital primary path coupled with a physically disjoint restoration path (Pongpaibool, 2002). This restoration path is pre-dtermined such that it speeds up restoration. During typical optical network operation, the IP channel is recognized along the service path only and resources are held in reserve along the restoration path. This reserved resouces on every optical link along any restoration path are mutual across various restoration paths whose service paths are not predicted to fail concurrently,such that these service paths do not use a common fiberspan. This is applicable since the restoration path is only recognized in case of failure in the service path. reserved resouces on the various restoration paths are dimensioned such that sufficient capacity is available in the recovery of all the IP links and optical network services that were affected during the occurrence of a single fiberspan failure.
3) Hybrid Architecture
In dimensioning the network in this architecture, the IP layer s first dimensioned with the associated DWDM layer resources vital in support of IP service traffic. Any supplementary IP router line cards together with optical network capacity necessary are determined that ensure recovery from single router or failures in the fiberspan. This is achieved by similarly considering each failure scenario, then determining the additional bandwidth requisites in recovery from failure. Given that all spare router line cards are shareable across independent failure occurrences, the quantity of spare router line cards necessary at each router iscalculated as the highest overall failure possibilities of the entire additional link channels essential for adjacent routers. All extra RON restoration capacity on every link is the utmost overall router failures. This can be calculated by determining the additional capacity needed by routers in recovery from the failure, while in fiberspan failures, the additional capacity is the necessary capacity for IP layer linkage and the private line services used in the RON. This capacity draws efficiency in this architecture from the adaptability of a single spare router line card in covering multiple, non-simultaneous instance of failures and the capability to allocate RON restoration capacity among IP and other services.
4) Modified Protected Architecture
In dimensioning the requisite network for this architecture, the IP layer resources are first dimensioned so as to support all IP service traffic. This is modelled through routing IP traffic on IP topology via the most express path algorithm. The extra IP spare router line cards together with the optical network capacity necessary is then determined in order to ensure recovery from all set of failures. This is again achieved through the enumeration of every failure possibility and then computing the additional bandwidth necessities in the IP layer and the RON attributed to each failure. Given that fiberspan failures are similarly restored in the optical layer, there is no need for a spare router line card for fiberspan failure possibilities. The spare router line cards are therefore implemented if failures occur in the router and router line card. In case of a failure in the router, the router line cards which have been connected at the remote ends of the IP links that are further linked to the failed router can be freed then used to re-establish fresh IP links in order to provide the much needed restoration capacity. Spare router line cards that are necessary at a router when failure occurs in another router can be computed as the total additional IP link capacities needed by adjacent routers after subtracting the total freed router interface cards as a result of failure. The least number of spare router line cards in every router is then calculated as the highest number of spare router line cards needed in the event of router failure over all router failures. A minimum of one spare router line card is assumed to be available on every router so as to aid in recovery from any single router line card failures shield (1:n interface protection, for large n). The extra RON capacity on every optical link emanates from the IP link along with the layer restoration capacity. This additional RON capacity used in IP layer restoration is computed as a maximum for every scenario of failure computed by determining the additional capacity needed by routers in failure recovery based on present restoration capacityavailed in the RON. The extra capacity for fiberspan failures is that required for recovery for the IP layer connectivity and all private line services carried over in the RON. The competence in capacity of the modified protected architecture is also enhanced by the fact that a single spare router line card is highly flexible in its coverage of multiple, non-simultaneous occurrences of failures and the capacity to share RON restoration capacity among IP and any additional services.