'Internet Protocol (IP) over DWDM' is the concept of sending data packets over an optical layer using DWDM for its capacity and other operations. In the modern day world, the optical layer has been supplemented with more functionality, which were once in the higher layers. This creates a vision of an all-optical network where all management is carried out in the photonic layer. The optical network is proposed to provide end-to-end services completely in the optical domain, without having to convert the signal to the electrical domain during transit.
Transmitting IP directly over DWDM has become a reality and is able to support bit-rates of OC-192. As we can clearly see, it holds the key to the bandwidth glut and opens the frontier of terabit Internets too.
[Back to Table of Contents]
The concepts of optical fiber transmission, loss control, packet switching, network topology and synchronization play a major role in deciding the throughput of the network. These factors have been discussed briefly in this chapter. More detailed explanations refer to
This block is also required to perform network management in the optical layer. The OXC, owing to its architecture, can easily perform Signal Monitoring, Provisioning and grooming, Restoration at the photonic layer itself.
All optical header replacement is the key to updating information in the wavelength-based packets (For example: Modifying routing information). In the case of a same-wavelength header, this is achieved by using a continuous wave (CW) tag attached at the beginning of the packet as carrier for the new header. This continuous wave tag does reduce the throughput of the network, but it had the advantage of maintaining spectral accuracy in the packet.
When the header is wavelength independent, replacement can be done at 1 Gbps, without using the static continuous wave tag that precedes the packet. The new header is created by optically modulating a continuous wave region generated from the data packet's own flag. This ensures the same wavelength for the new header, as in the original packet. This continuous wave tag generation adds certain complexities to the packet, but the overall packet structure is maintained.
There exists the concept of a virtual topology that indicates the lightpaths from one node to another. This arises from a logical view all connections in the network (and not the physical structure). The virtual topology will comprise N x N connection. Hence, determining a light path constitutes an N 2 -scaling problem. An efficient algorithm needs to be devised to calculate an optimal virtual topology based on the traffic pattern. There exist many algorithms for dynamically reconfiguring the virtual topology of the system. In most cases, knowing the physical structure would help reducing overheads, by keeping the number of fiber links traversed to a minimum.
A ring topology may be preferable in most cases, owing to many of its capabilities. Unlike a mesh network, the expense of laying out the links is reduced in the ring, because the number of links increases only as a linear progression. The ring topology besides serving as a standby link helps share the load. The working segment (Refer to Fig.2) and the protection segment of the fiber together handle the large data burst of the computer network. This reduces the load on the router and removes the need for buffering. Nonetheless, mesh networks provide a faster restoration in the system.
Fig.2 Ring Topology Connecting Nodes A & B
Nevertheless, synchronization may still be used for assuring good transmission quality. The numerous regenerators / transponder and other devices in the path of an optical signal introduces jitter. Synchronization can be used to ensure quality by cleaning up the signals transmitted at each node. SONET terminals and ADMs have a special timing output port, which provides timing to customers. It is sometimes referred to as the Derived DS1. It is a true DS1 signal, but carries no traffic. All data bits are set to logic 1 to minimize timing jitter. A clock distribution amplifier may be used to split the Derived DS1 signal, to synchronize many network elements. In a network, each distribution amplifier output may be routed to a different network element.
[Back to Table of Contents]
Fig.3 Primitive Protocol Stack for SONET over DWDM
This multi-layer stack is required to maintain a division of labor. The ATM layer is used as an access technology and might be limited to a speed of OC-12. The IP layer acts as the data plane in this case and operates at less that OC-3 speeds. The system performs all important tasks like signal monitoring, provisioning and grooming, restoration.
The multi-layer stack has more problems than advantages. Functional overlap is one such problem. Each layer tries to perform restoration in the event of a failure, thereby creating more havoc in the system. The SONET interface is advantageous for constant bit rate traffic, but not for bursty traffics found in the Internet. Presence of high capacity in the system obviates the need for time division multiplexing and traffic grooming. Thus a multi-layer stack introduces undesired latency. The SONET system does not provide fast provisioning. It provides protection in the form of redundancy. (One more fiber cable in the form of a ring). All these problems have made it necessary to redesign the system without having too many layers.
The other proposed solution for transmitting IP packets reliably over a fiber optic network is to adopt an IP over DWDM system. The transport layer is all-optical and thereby maintains a high data rate. This is more favorable because it avoids the cost of the SONET / ATM equipment. SONET does not work very well with bursty traffic. The need for SONET's multiplexing is no more present.
The need for the IP directly over DWDM system has set in to increase the bandwidth and to reduce the latency. The system performs satisfactorily at high speeds of OC-192. The overheads because of the SONET & ATM layers have been eliminated. The new architecture facilitates for faster restoration, provisioning & path determination. It seems to have much potential as the architecture of the future.
The optical transport architecture will employ both transport networking and enhanced service layers, working together in a complementary and interoperable fashion. The functionality in the optical layer can be split as given in Table1. The two layers perform the functions of the four SONET layers. The transport layer acts like the lower physical layer, while the service layer acts like a higher layer.
Table 1. DWDM Network model
Wavelength level traffic-control
|Access speed, Usage rates, Security, VoIP services etc|
The two layers together achieve the granularity needed by all the services (like traffic engineering). In this model, the SONET gives way to the optical transport, which tries to achieve reliability and performance as provided by the standard SONET architecture. Having an intelligent optical layer that performs fast restoration can appease the bandwidth demand. Restoration happens in the optical layer rapidly and does not overlap with the service layer's mechanisms. Switching & bandwidth is furnished at the granularity of the wavelength. The ATM's virtual path becomes equivalent to a wavelength. Furthermore, the Multiprotocol Label Switching (MPLS) protocol divides the traffic engineering requirements between the IP layer and the Optical transport layer. Thus a DWDM layer with required functionality is molded to form the all-optical network.
Achieving the benchmark set by SONET leads to many complex expectations. When a connection is established, the DWDM layer has to provide both reliable optical path and traffic engineering. An automated provisioning of end-to-end wavelength path should happen in a virtually less amount of time. In case of a physical failure, the wavelength routing protocol will have to restore the transcontinental paths across many routers within a maximum of 50 ms. These expectations bring about a vagueness in the extent to which some functionality should be pushed down to the optical layer.
Fig 4: DWDM Network Architecture
The closed architecture was designed to serve the SONET systems better. It increases the capacity in the SONET system, by utilizing the necessary components and the technology of DWDM in the standard SONET terminal. It is dependent on the higher SONET layers or any other TDM system for its other functionalities (like Network management). The segment A in Fig.4 denotes such a system. Here, the carrier gets stuck with the vendor's proprietary technology.
IP/DWDM systems adopted the alternative approach, which yields a whole new transport layer, called the Open architecture. It is open in the sense that it is not tied with SONET or other TDM systems. This case reflects protocol transparency and exhibits all the properties of the all-optical network. For optical networking to realize its full potential there must be a standard interface to the optical layer. The segment B in Fig.4 denotes the open system. The customer is responsible for providing the actual interface to the end user and for all the protection work. The IP bits enter the DWDM system and then are transported "as is" over the high-speed connection.
The IP/DWDM system can adopt a variety of architectures based on what the payload is and what the underlying transport network is. [Doshi98]. They can be grouped as:
Fig.5 Wavelength transport network
SONET serves two basic functions: multiplexing and network restoration. If the network survivability of the DWDM system improves, then SONET will become obsolete. SONET's reliability represents wastage by setting aside a large portion of its resources. But, it is the cheapest solution available for incorporating reliability in the circuit-based system. This it does by supporting a four-layer architecture. Moreover, SONET solutions are very much vendor specific. The advent of gigabit routers makes the need for SONETs even less because the increased capacity seems more than sufficient.
In contrast, the IP over DWDM system has inherent advantages because of the absence of many layers, thereby cutting down on the cost part. The signals need not be converted onto an electrical domain for performing control operations on it. Hence, the latency in the IP/DWDM system is less compared to that encountered in the SONET system. The absence of a vendor specific component makes the system service transparent.
The result of this combat between SONET & IP/DWDM ends in a tie. The primary reason being that the current implementation of IP over photons does not exist without a SONET interface. Many carriers and vendors predict that SONET will still be around for a while. Many others predict that the features of SONET shall soon be burnt into the IP/DWDM system, leading to the end of SONET.
[Back to Table of Contents]
In the case of the other non-SONET data, the system becomes a bit more complex. Transporting signals over the DWDM layer directly enhance protocol transparency. But, it hampers bit-error checking. Therefore, fault detection is hindered and the maximum distance traversed without possibility of a bit-error reduces.
Forward Error Correction
Forward error correction (FEC) is performed in the all-optical DWDM systems. It can be categorized into two types. The first way is to put in the FEC data onto the unused portion of the SONET overhead. The performance is limited by the fact that SONET frame has a restricted amount of space available in its frame. This is also known as the in-band FEC. The other alternative is to have the FEC data encoded and transmitted on the line separately. This method, also known as the out-band FEC,increases the line rate and hence, provides significant system improvement.
Optical Cross Connects (OXC) can be used as an integral part of this protection architecture. It can provide a 1+1-protection scheme via a head-end bridge, while the tail-end OXC can be provisioned to flexibly switch between two receive optical ports, based on signal quality. This 1+1 optical layer protection switch guards against fiber cuts at the highest possible level, which is architecturally the most appropriate solution. Fig.6 depicts the topology consideration during restoration.
The strategies for network restoration and survivability will undergo significant advances in the years to come. MPLS holds out good scope for such advancement, by allowing optical networks to carry out the restoral and path protection switching at the IP layer, rather than at the photonic layer. Furthermore, Optical networking will, in time, support powerful self-healing capabilities consistent with those of SONET/ SDH needed in a comprehensive optical layer.
WADM can interconnect the IP routers, making it possible to establish a light path between the two. (A light path is the path an optical signal traverses to reach the destination beginning from the unique source. During transit, it may pass through wavelength converters. But, a wavelength path is the light path devoid of wavelength changers). The routers essentially reduce to neighbors thereby changing the network topology perceived by all the participating IP routers. This type of flexibility implies that the topology assumed by the IP routing protocols, could be changed as traffic conditions vary.
There are two solutions to the routing problem [Anderson99]:
Some of the issues that must be faced to deploy IP gigabit backbones are:
1. Outlining the specification and dimension of the network and network elements
2. Evaluating the impact of customer bandwidth on network architectures
3. Using network design and operation principles.
4. Evaluating the role of all-optical end-to-end paths
5. Evaluating the role in the routing and wavelength assignment problem
6. Design the wavelength pool and performing wavelength reuse.
7. Introducing integrated network management for the backbone access.
These are yet to be incorporated satisfactorily in a IP/DWDM system. Many products achieve this by providing a separate network management system that has full control over the complete optical network.
Bit rate independence is a necessary condition for being service transparent. The optoelectronic processing in a network can introduce minor jitter. To remove the jitter and to regenerate a good quality signal, certain timing relations need to be used. This hampers the bit-rate independence. This dilemma can be resolved by using a bit-rate independent optoelectronic regenerator with re-timing functionality. [Alferness99]
The protocol transparent and bit-rate independence together establish the service transparency, which is essential for developing an all-optical network and a complete optical transport layer.
According to a Juniper press release, the interoperability between the Juniper Networks M40 Internet backbone router and CIENA's MultiWave Sentry
DWDM optical-networking system has been successfully tested. This announcement demonstrates the direct connectivity of Juniper Networks' and CIENA's equipment and indicates the availability of an integrated IP over DWDM solution that simplifies IP packet transmission via an optical core.
The above consideration requires analysing as to what benifits are reaped by performing traffic management at the optical layer (instead of at the IP layer). Moreover some thought has to be given in determing whether those functions can be implemented effectively in the optical layer. In most cases, the IP layer can be modified to take care of the routing based on the loading conditions and the optimal path. This avoids functional overlap and thereby improves the system performance. An analytical study is hence needed to decide between a QoS based distributed routing scheme in the IP layer and an optimal routing algorithm undertaking the IP/DWDM routing.
[Back to Table of Contents]
IP/DWDM system tries to merge into one layer the functionalities of the ATM switch, SONET Mux / Demux and IP routing. This feature of the optical layer has brought in the concept of a Multiprotocol Lambda Switching protocol, which will perform those operations. The protocol shall try to implement routing protocols, enforce QoS & perform protection. This section discusses the features & architecture of the protocol adopted.
[Awduche99]describes much of the work in this area.
A DWDM network is analogical to an ATM network in the aspects of switching. ATM networks perform packet switching based on the virtual circuit number, while the optical channel layer performs switching based on the wavelength of the signal (or packet). Hence the name "Lambda Switching" is applied to the optical network.
An Optical Channel trail is the complete sequence of wavelengths assumed by the Ip packet in transit. This sequence is similar to the stack of labels in MPLS. If the node did not contain a wavelength converter, then the circuit identifier (in this case a wavelength bound by the operational bandwidth) has global significance. Cross connects contain different ports each characterized by a unique wavelength. It is reponsible for switching the packets appropriately.
The Mulitprotocol Lambda Switching currently has many limitations. It does not support TDM. There exists no buffering of packets. Hence no scheduling algorithm is required. The packets are sent as and when they arrive. The large capacity of the fiber facilitates such an operation. The implementation of the Lambda switching, currently, is vendor specific. Thereby hindering interoperability between the systems.
The real implementation should have an integrated control plane. Each component (like OXC, Label Switching Routers) should not have different control plane. The drawback in IP over ATM, where both IP and ATM has distinct control planes, shows us that such a control plane is disadvantageous. The components shall have an uniform control plane. The MPLS traffic engineering control plane would be suitable for that of OXCs. The OXC using that control plane will be an IP addressable device. Thus a new architecture for the MPLS control plane is born.
Fig.7 Network Architecture of the future
The required protocol stack, for achieving those goals, would be as shown in Fig.7. The MPLS control plane shall regulate all the connections. The switch fabric would perform packet switching based on the wavelengths. The data plane shall be responsible for transmitting the packets. The switch shall maintain mapping between the wavelength + port at input and the wavelength + port at output. The table ensures that the packets reach the proper destination eventually. This is similar to how wavelength routing is done.
[Back to Table of Contents]
The future holds many a challenges to the all-optical networks. But, the commercial implementations for IP over DWDM are not far away. It opens the pathway to Terabit networking and unleashes the enormous bandwidth potential of the silica fiber. The trend of IP/DWDM solutions over the last few years seems to have taken an exponential growth. DWDM acts as the stepping stone towards a true optical networking era.
[Back to Table of Contents]
[Awduche99]D. Awduche, Y. Rekhter, "Multi-Protocol Lambda Switching: Combining MPLS Traffic Engineering Control With Optical Crossconnects", Internet Draft, Work in Progress, 17 pages, 1999.
Cites recent work in the design of Multiprotocol Lambda Switching for supporting QoS.
[Chen99] Chen, Yi et al, "Metro Optical Networking", Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.163-186.
Detailed explanation of the optical layer issues, pertaining to metro-networks.
[Alferness99]Alferness, Rod et al, "A Practical Vision for Optical Transport Networking", Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.3 - 17.
A practical view point to the optical layer issues.
[Anderson99]J.Anderson, J.Manchester, A.Rodriguez-Moral, M.Veeraraghavan, "Protocols and Architectures for IP Optical Networking", Bell Labs Technical Journal, Jan-Mar 1999, pp.105 - 124.
The design & implementation strategies for the various issues involved.
[Wilner99]Wilner, Alan, "Key limitations in WDM systems and networks", Conference Proceedings, SPIE, Vol. CR 71, pp. 220-245, 1999,
Technical explanation to the loss compensation and header replacement methodologies.
[Mukerjee97] Mukerjee, Biswanath, "Optical Communication Networks", McGrawHill, July 1997, 575 pages, http://networks.cs.ucdavis.edu/users/mukherje/book/toc.html.
Explains various concepts with WDM systems and algorithms for packet routing.
[Wei99] L.Wei, Y.Chen, G.Wong, "The Evolution of China's Optical Fiber Networks", Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.125 - 145.
Explanation of strategies used in practical situations.
[Lucent99] Lucent Technologies , "Web ProForum tutorial: DWDM", Oct 1999, 15 pages,
General overview of DWDM systems and about transmitting IP over it.
[Alcatel99] Alcatel, "Web ProForum tutorial: Optical Networks",Aug 1999, 29 pages,
Brief description of various issues involved in optical networks.
[Giles99] Giles, Randy and Spector, Magaly, "The Wavelength Add/Drop Multiplexer for Lightwave Communication Networks", Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.207 - 229.
Explains the architecture and implementation details of WADMs.
[Jackman99] N.Jackman, S.Patel, B.Mikkelsen, S.Korotky, "Optical Cross Connects for Optical Networking", Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.262 - 282.
Explains the architecture and implementation details of OXCs.
[Refi99]J.Refi, "Optical Fibers for Optical Networking",Bell Labs Technical Journal, Vol 4, No 1, Jan-Mar 1999, pp.247 - 281.
The technical details of the optical fiber have been discussed in detail.
[Back to Table of Contents]
Dense Wavelength Division Multiplexing
Forward error correction
International Telecommunications Union
Multiprotocol Label Switching
Optical Cross Connect
Quality of Service
Synchronous Digital Hierarchy
Synchronous Optical Network
Time Division Multiplexing
Wavelength Add/Drop Multiplexer
Wavelength Division Multiplexing
[Back to Table of Contents]
Note: This paper is available on-line at http://www.cse.wustl.edu/~jain/cis788-99/ip_dwdm/index.html