Mesh vs. Ring: More than just network efficiency

The current economic climate of the telecom industry may fool us into thinking that the sharp reduction in capex would suggest a similar reduction in opex, which includes the expenses for operating the network up to the level of the customers' satisfaction. However, even under the pressure of a very competitive market--where the average price for bandwidth drops approximately 30% to 35% yearly--carriers still state that they are nowhere near ready to sacrifice their network availability in order to improve profits. Even though one might see capex and opex as two different animals, in fact they relate to each other to a great extent.

Therefore, the main focus should be controlling the cost as a whole, not individually decreasing capex and/or opex with an expectation of reducing total cost as a result. It might be possible that the solution which gives only the lowest capex has very low availability, such that either the network performance does not make economic sense to the customer in terms of its needs and satisfaction or opex cost may increase significantly in order to bring the availability to a reasonable value by an attempt to shorten the Mean Time To Repair (MTTR) with an ultimate increase in repair crew efforts. To avoid such pitfalls, it is required to bring availability awareness into the modeling exercise of various network architectures, as each network architecture might provide different levels of availability.

The critical debate in architectural solutions has always been between mesh and ring. In the last several years there has been growing interest in mesh network topologies. This has been fueled by vendors claiming that mesh networks have much better efficiencies and, therefore, save money for carriers. However, mesh has been always under scrutiny for its restoration characteristics. But new protocols based on data networks can bring rapid distributed mesh restoration without the need for a centralized controller; therefore, these new protocols offer almost the speed of ring restoration with the efficiencies of mesh.

On the other hand, there have been efforts to reduce the cost of mesh architectures. One of these efforts is the introduction of express routes to a mesh architecture. In this new architecture, express routes are defined between specific nodes that will carry large amounts of traffic bypassing intermediate nodes. On average, this architecture requires more transmission systems and more fiber than the traditional mesh architecture--the aim is to reduce the amount of equipment at the nodes by trading off additional transmission systems.

Node sizes will be smaller on average, and therefore so will be the node cost. From a cost perspective, the design of this architecture becomes a compromise between the transmission system cost and the node cost. When we look at the other side of the coin, it is also a question of how this increase in transmission systems and decrease in node complexity and size will affect the overall availability.

Ring architectures are known to provide better resiliency at the cost of lower efficiency in terms of network utilization in comparison to meshes. Today, Bi-directional Line Switched Rings (BLSR) represents the most deployed architecture in backbone ring networks. Ring architectures allow a fast and simple restoration mechanism compared to the complexity of mesh restoration. Restoration bandwidth is allocated in each ring in such a way that any single failure is resolved in the very same ring, without any effect on the rest of the network--i.e., ring failure isolation. Taking all that into account, we can conclude that ring architectures are usually more robust than meshes; however, mesh architectures seem to be more economical than ring architectures.

Line Shared Protection Rings (LSPR) is an extension of BLSR. LSPR combines the benefit of the fast restoration of the ring architectures and the economics of the mesh architectures. The differentiating characteristic is its sharing of the restoration bandwidth among adjacent rings. The restoration bandwidth in a link shared by several rings and can be used to protect against failure in only one of these rings at a time.

Availability is a very important factor that is often underestimated in network design based on the assumption that the network will meet the needs of customers, which has always been five 9's of availability in backbone networks. However, with today's ever-enlarging spectrum of services requiring different levels of availability, availability needs to be carefully engineered. Availability is rarely intuitive--for example, a ring network isolates restoration to a single ring, which allows multiple failures to be restored simultaneously if they occur in different rings, but ring networks use more fiber and equipment than a well-designed mesh. More fiber and equipment means higher failure rates, and therefore lower availability, but the ability to restore multiple failures simultaneously means better availability.

If all components were at five 9's, the network availability becomes significantly less than this; therefore, to reach a target availability value of five 9's for a network, the component availability in reality must be much better than five 9's. So then, the question becomes "What is the crucial component to improve overall network availability?" Is it fiber? Is it amplifiers? Or is it the node equipment? According to what we learned from a sensitivity study, the target network value of five 9's can only be reached with nodes having significantly better than five 9's availability, even if the availability of either transmission systems or fiber is much greater than five 9's.

In an effort to decide which network architecture to adopt for a typical U.S. network shown in Figure 1, we used some generic availability values.

For fiber, we used two failures per year per 1600 km with 4hr. MTTR. For nodes with protection, we used eight 9's. We used 0.9996 for regenerators (unprotected) and 0.99997 for in-line amplifiers. With these numbers, Figure 2 plots end-to-end connection availability for every channel in the considered architectures: BLSR, LSPR, Mesh, and Mesh with express routes.

Cost comparisons, where generic equipment costs are considered, are given in Figure 3.

If we look at the results in terms of the absolute dollar amount, mesh with express routes has the lowest cost. Express routes bypassing nodes, significantly lowers the dominant OXC port cost at the expense of an increase in fiber length. This increase in fiber length does not lead to a significant cost increase, because the existing conduits and cables are considered for deploying the excess fiber and therefore, no additional construction related costs need to be accounted for. With the increased fiber length of express routes, the availability drops in general due to a higher likelihood of fiber failure and a larger number of amplifiers and regenerators. The average availability drops from 0.999997411 to 0.999996841. In addition, the end-to-end availability of more channels drops below the threshold level of five 9's; it increases the number of channels below the five 9's threshold to 9.07% of all the channels compared to 5.33% of channels below threshold for a mesh without express routes.

Here is the question: is the service provider a fiber rich operator and can it live with 9.07% of the channels below-threshold? If so, then the mesh with express routes is the best architecture as it can save 36% of total cost compared to a normal mesh, see Figure 3. On the other hand, if deploying fiber in a service provider's network requires a significant investment in terms of construction and labor and/or 9.07% of the channels below threshold is not acceptable, then service provider is better of with a mesh architecture without express routes or a ring architecture. The BLSR architecture has the highest cost, as can be seen in Figure 3, but its average channel availability is better than all other architectures with only 3.16% of its channels below the threshold. In addition, the availability of the worst channel for the BLSR is much better than the availability of the worst channel of any of the other architectures. Assuming that the operator requires all channels above the threshold, the costs of fixing 3.16% of the channels in BLSR versus 5.33% in mesh needs to be considered in the decision of which architecture makes the most sense economically. It is possible to reach mesh architecture efficiency with the LSPR architecture, due to its shared restoration capacity concept. It saves more than $180K on port cost, $20K in amplifier cost, and $7K in fiber cost compared to the BLSR.

On the other hand, the number of channels below the threshold value is very similar between the LSPR and BLSR architectures. Therefore, LSPR is a good candidate as it is close to mesh efficiency while still having ring restoration speed. Availability wise, it is also very close to the BLSR. LSPR and mesh are comparable architectures from a cost point of view. They have almost the same number of OXC ports, fiber and amplifiers, but the mesh has more paths below the threshold value. Therefore, LSPR seems to be a very promising architectural candidate for both cost and availability, assuming that deployment of additional fiber for a mesh with express routes is not cost effective. In some cases the service provider cannot afford to have any channels below the availability threshold; therefore, a solution that may be more costly but has the least number of channels below the threshold that could be individually improved would be the right solution. Therefore, the optimal architecture for a network involves more than capex and/or efficiency considerations--it depends on the service provider's needs, what it has in its network, and the quality of services it wishes to offer.

Contact Gary Ester at gary.ester@alcatel.com, Hakki Cankaya at hakki.cankaya@alcatel.com and Ana Lardies at ana.lardies@alcatel.com