Do operators need optical core (DWDM) networks using 40Gbit/s channels, and when and under what circumstances should 40Gbit/s systems be deployed? Siemens gives a different perspective on the economics of the transition.
By Marc Sauter and Wolfgang Sitter, Siemens ICN
With telecommunications carriers faced with both overcapacity in the industry and an economy that is struggling, is it reasonable at this time to discuss increasing capacity?
Apparently so. Recent market research reports indicate that the annual duplication of Internet-driven traffic is continuing [1], a fact perhaps drowned out by the current cries of economic panic. Indeed, part of the restructuring opportunity that this downturn is providing involves the need for continued investment in networks in order to remain competitive, i.e. reducing operating costs and increasing revenues in response to increased traffic volatility.
Is 40Gbit/s the answer?
The benefits of operating a core network at transmission rates of 40Gbit/s come mainly from the reduced number of optical components needed in core switches. The size and number of switching matrices and the number of filters and channel amplifiers is reduced, since only about a quarter of the channels are needed.
Smaller systems need less floor space and power, as well as administrative effort (reducing operating expenditure).
Past bandwidth-driven network build-out has been superseded by the quest for improved cost efficiency. For certain network configurations and where there is much traffic, the total cost of 40Gbit/s networks can be less than for 10Gbit/s networks, even with the optical component cost forecast for 2002/2003.
Networks operating at 40Gbit/s can also deliver a higher total throughput per system and per fibre. This might not be needed now, but will be in the near future [1, 3]. Also, by establishing or upgrading the network in a modular fashion, starting with a small number of 40Gbit/s channels, there are immediate benefits through cost savings. Later, the network can be expanded as required, enabling operators to respond rapidly to demand for more bandwidth.
Challenges
The transition of fibre-optic infrastructure from 10Gbit/s to 40Gbit/s data rates involves technical challenges from both linear and non-linear optical effects.
The energy of the 40Gbit/s light pulse is only a quarter that of the 10Gbit/s pulse, requiring very sensitive receivers. Simply increasing power, however, causes other undesirable effects like four-wave mixing.
The physical challenges apply to fibres and switching elements alike.Also, 40Gbit/s systems are much more sensitive to signal distortion from photonic switching elements, so unregenerated distances are restricted and additional OEO regenerations can increase total system cost.
Another challenge is to fill the 40Gbit/s channels with traffic of lower client speed. Since no 40Gbit/s client signals will exist for some time [2], all signals have to be multiplexed into 40Gbit/s channels. Depending on the end-to-end connectivity, the number of low-speed client channels and the method of multiplexing them, a certain amount of over-provisioning is generated.
With intelligent grooming, this can be reduced. But, for an economic network configuration, the operator must balance the additional planning effort needed for grooming and the cost savings due to grooming.
Optimised system design
Our economic network analysis of 40Gbit/s versus 10Gbit/s technology is based on a core network of a typical national carrier in Europe.
The best economic evolution of a transport network is achieved when upgrading the network from 10Gbit/s to 40Gbit/s stepwise, i.e. it is prudent to upgrade those channels to 40Gbit/s where required and where network topology permits; all other channels should remain at 10Gbit/s until economic conditions allow. Such a stepwise evolution to 40Gbit/s requires DWDM systems, which support parallel operation of 10Gbit/s and 40Gbit/s channels.
The following are from a study of the economic viability of an upgrade in various permutations:
- DWDM-fibre switch-DWDM architecture: opaque solution where all traffic is terminated at every node. 40Gbit/s channels are de-multiplexed, optically switched at the 10Gbit/s level, re-groomed to improve the economics and multiplexed into 40Gbit/s channels.
- DWDM-WSXC-DWDM architecture: an all-optical network solution with wavelength-selective photonic cross-connect where express traffic is not electrically processed and transponders are only used for add/drop traffic entering or leaving the core network.
- DWDM-OADM-DWDM architecture: all-optical network solution with photonic add/drop device, where express traffic is not electrically processed and, as in the WSXC node, transponders are only used for add/drop traffic entering or leaving the core network.
Deployment of these architectures in two specific sub-networks of the core result in four scenarios yielding various advantages:
- Scenario1: Opaque meshed point-to-point (P2P) network switched by node type (1), DWDM-fibre switch-DWDM
- Scenario2: All-optical meshed network switched by node type (2), WSXC
- Scenario3: Opaque linear chain-network switched by node type (1), DWDM-fibre switch-DWDM
- Scenario4: All-optical linear chain-network switched by node type (3), OADM.
In an economic comparison of the different scenarios, opaque 10Gbit/s P2P is used as the reference value (i.e. 100%). To see the cost structure of the different solutions seven component classes are shown in Figures 1 and 2: DWDM terminal, switching element, optical line repeaters, 10Gbit/s transponders and regenerators, and 40Gbit/s transponders and regenerators.
Fig. 1 shows the results of comparing scenarios 1 and 2 at 40Gbit/s with their respective scenarios at 10Gbit/s for traffic demand in 2002. These are deployed in the heavily meshed sub-network - the node type is (as appropriate for meshed networks) the photonic cross-connect node WSXC.
At current traffic volume, both the P2P and WSXC 40Gbit/s configurations are more expensive than their 10Gbit/s counterparts. There are no savings for WDM terminals due to low-channel-count systems. So, the higher cost for 40Gbit/s regenerators and transponders cannot be compensated for. There is, however, a cost benefit of about 10% when deploying the photonic WSXC instead of the fibre switch at 10Gbit/s.
The configuration most suitable for deployment of OADMs is a subnet composed of a linear chain of nodes and links. The comparison of OADM vs P2P for 10 and 40Gbit/s for traffic demand in 2002 is shown in Fig. 1(b).To determine how cost of scenarios varies when traffic load rises, network traffic was changed to a level expected for 2004 - four times that for 2002 - based on annual duplication of Internet traffic [1]. Figure 2 shows the cost at the high traffic volume for both the meshed and the linear chain sub-network.
With increasing network load, the 40Gbit/s network quickly becomes competitive with the 10Gbit/s scenarios. Both deployments of photonic switches (WSXC or OADM) at 10 or 40Gbit/s are clearly beneficial compared to cost of 10Gbit/s point-to-point configuration.
In the mesh network case, cost at 40Gbit/s is lower due to the reduced number of channels and lower cost of DWDM equipment. This is only partially offset by higher cost for OEO, resulting in net cost savings.
In the linear chain network, photonic OADMs can be used. So, the good long-haul performance of the 10Gbit/s system is decisive for the highest cost benefit, giving cost savings of about 40% for the 10Gbit/s OADM configuration. The 40Gbit/s systems are still more economic than the 10Gbit/s P2P, but the 40Gbit/s OADM yields only 15% cost benefit, since it needs more OEO regenerations and is therefore more expensive than the 10Gbit/s OADM solution.
ConclusionWith this economic evaluation we have examined the capital expenditures (CapEx) of 40Gbit/s networks and compared these to identical networks operating at 10Gbit/s. The cost of 40Gbit/s equipment was estimated for 2002/2003. We have shown that, for year-2002 traffic volume, networks operating at channel speed of 10Gbit/s are more economic than 40Gbit/s networks for both linear chain and meshed networks. A photonic OADM device yields the lowest capital expenditure, but cannot be deployed in meshed networks. In meshed networks, a cross-connect is needed, a photonic switching element helps saving cost, as we have shown.Assuming that network traffic will continue to double annually - depending on the geographic region [1] - 40Gbit/s networks will become more economic than 10Gbit/s networks: in 2004 in a mesh environment a 40Gbit/s network is more economic than the 10Gbit/s network, even though we use 2002/2003 estimated 40Gbit/s element costs.
Due to the high cost of 40Gbit/s optical components, only the combination of the benefits of 40Gbit/s and all-optical network solutions grant an economic advantage of 40Gbit/s over 10Gbit/s networks. Also, network topology, end-to-end connectivity and network traffic load have a major impact. As expected, 40Gbit/s networks may have a slightly higher entrance cost at low network load, and become cost efficient at higher network traffic.
The best economic advantage of 40Gbit/s networks can be achieved when deployed in highly meshed networks with high proportions of direct point-to-point connections between nodes, i.e. little through-traffic (short transparent lightpaths). Cost benefits may be further increased by intelligent grooming through advanced network planning and engineering, which reduces the over-provisioning tax.
The savings from introducing 40Gbit/s channels into the core network can be as much as 22%, depending on the implementation (e.g. OADM or WSXC), network size and connectivity, and on traffic demand pattern. OADMs are a perfect fit at intermediate sites (i.e. only two fibre pairs) [4], whereas WSXCs mainly appear at major sites with high connectivity (e.g. ring interconnect or mesh networks).
A core network operating at 10Gbit/s can be economically extended by introducing 40Gbit/s channels into the network and taking advantage of optically transparent switches. The upgrade should be done by converting only those 10Gbit/s channels into 40Gbit/s channels for which traffic volume is justified. Such an evolution requires DWDM systems, capable of operating at both 10Gbit/s and 40Gbit/s channels on the same fibre. 10Gbit/s channels are used for long transparent connections where its span performance is necessary, and 40Gbit/s channels are deployed for short direct connections between two neighboring nodes where the capacity is required.
Siemens provides such systems for a smooth network evolution, and also has efficient multiplexers in its portfolio which serve the purpose of aggregation and grooming of lower-speed client signals into 40Gbit/s channels.
To conclude, networks at 40Gbit/s deploying cross-connects and OADMs with photonic switching elements allow the carrier to build highly scalable and, most importantly, cost-effective future-proof core networks. The path of network evolution will be gradual, by moving from 2.540Gbit/s to 1040Gbit/s then 40Gbit/s, upgrading 10Gbit/s channels to 40Gbit/s, and increasing the channel count for a total increase of transported traffic.
With the expectation of further increase of network traffic and falling cost for 40Gbit/s equipment, CapEx may fall even more compared to 10Gbit/s. Less equipment and fewer channels to operate also means a reduction of operational expenses per bit/sec.
Wolfgang Sitter
Product Manager
Siemens Optical Networking Division
E-mail: [email protected]
Sitter joined Siemens in Munich as a SW development engineer for communication products and has worked in various positions for Siemens developing and marketing products for data communication networking.