IP/optical control plane integration typically results in cost savings. But for a good return on investment (ROI), it is also crucial to maintain service reliability and availability. So how can service providers achieve these goals simultaneously?
Using multi-layer protection and restoration is the most successful approach, according to a Nokia Bell Labs analysis of three modes of operation. It offered a far more cost-effective way to maintain availability — and delivered TCO reductions of up to 40 percent, with accelerated ROI.
A better option to improve efficiency
Running a network hotter may help maximize return on network investments. But when admitting too much traffic, increasing network congestion may degrade service availability, and lead to cost penalties that can easily undo any potential savings.
Traditional 1+1 optical network protection maintains service availability by keeping 50 percent of network capacity in reserve. The alternative approach — only leveraging MPLS-based protection and restoration mechanisms at the routing layer — is equally inefficient in its use of network resources, although these inefficiencies are less immediately apparent. Nevertheless, in many networks this limited use of IP and optical networking capabilities is the present mode of operation.
However, better alternatives are available today, using state-of-the-art optical transport networks andreconfigurable optical add/drop multiplexer (ROADM) technologies. Agile optical transport with an intelligent control plane can provide transport layer resiliency, along with cost-effective utilization of networks resources.
The protection capabilities of these networks leverage crucial advances in the generalized multiprotocol label switching (GMPLS - RFC 3945) architecture. GMPLS adopts key concepts from the MPLS control plane used in IP routing, but adds functional enhancements to support multi-layer optical transport networks.
Figure 1. GMPLS protection and restoration options for the optical layer
As a result, service providers can use GMPLS to drastically expand their existing toolkit of network traffic protection and restoration capabilities, as shown in Figure 1 on the left. With the right architecture, GMPLS-based transport layer recovery can be combined with protection mechanisms in the IP/MPLS routing layer. Adding these capabilities allows service providers to offer a range of differentiated service level agreements (SLAs) that address availability needs for different classes of service, as shown in Figure 1, right.
SLA requirements can then be mapped on an appropriate multi-layer traffic protection and restoration strategy to balance availability, redundancy, and resource utilization for the best returns on network investments.
Which resiliency mechanisms are best for minimizing costs?
To get specific answers about cost effectiveness, Nokia Bell Labs compared traditional MPLS and new GMPLS-based resiliency mechanisms for the optical transport layers.
The TCO analysis used a backbone reference network model with six core routing nodes and five optical transport nodes. The physical transport network topology was partially meshed, while the core routing topology was a logical mesh.
The study compared network resource requirements for a mixed traffic matrix, with traffic growing evenly at 40 percent annually over a 5-year study period:
- 10 percent was expedited forwarding (EF) traffic. This is typically “lifeline traffic,” such as VoIP, that is sensitive to delay, jitter, packet loss, and outages. Protecting and restoring EF traffic has the highest priority. It must be resilient to multiple failures with restoration within 50 msec.
- 30 percent was assured forwarding (AF) traffic. This mission-critical, high-revenue data traffic requires reliable transport. But it can compensate for limited packet loss, for instance through retransmission by the transmission control protocol (TCP). It must be resilient against a single failure with restoration within 500 msec.
- 60 percent was load-balanced best effort (BE) traffic. This is mostly high-volume internet traffic with low revenue per bit. It has the most relaxed availability requirements. Nevertheless, service providers want to prevent long and frequent outages, and 1+N redundancy can provide resiliency against a single failure without nominal capacity loss.
The network protection and restoration strategies
The study compared the following three strategies, which are also illustrated in Figure 2:
- Leverage MPLS to protect and restore all service traffic at the routing layer, which is the present mode of operation for many networks
- Leverage GMPLS to protect all IP traffic at the photonic switching layer, the study’s first “future mode of operation”
- Leverage both MPLS and GMPLS/UNI in a multi-layer resiliency scheme, the study’s second future mode of operation
Figure 2. Network service protection and restoration strategies for IP over DWDM
Present mode of operation (PMO)
The PMO strategy protected against capacity degradation when a single LSP or optical link segment fails. It addressed each type of traffic in the following ways:
- It applied 1+1 redundant label-switched paths (LSPs) with MPLS fast reroute (FRR) over unprotected, but physically disjoint transport links to protect EF traffic end-to-end against multiple failures, with very fast restoration times below 50 msec.
- AF traffic was carried by non-redundant LSPs with MPLS FRR over unprotected, but physically disjoint transport links.
- For BE traffic, it used N+1 unprotected, physically disjoint LSPs in an equal-cost multi-path (ECMP) load-sharing model.
Future mode of operation 1 (FMO1)
The FMO1 strategy applied optical segments with GMPLS 1+1 protection and restoration combined to protect IP overlay traffic. This approach protected EF traffic against multiple failures with a restoration time below 50 msec. It differed from the PMO by making service restoration transparent to the IP layer and acting on aggregated traffic in the optical transport layer.
Future mode of operation 2 (FMO2)
The FMO2 strategy applied a combination of MPLS FRR over optical transport links with GMPLS guaranteed restoration. This method protected both EF and AF traffic against multiple failures, with rapid protection switching within 50 msec.
Wavelength restoration times at the photonic switching layer are on the order of seconds. So MPLS FRR was used in FMO2 to provide rapid restoration over alternate optical segments while the failed primary segment was being restored by GMPLS. Optical segments were able to share spare resources for restoration purposes. During FRR restoration, full bandwidth recovery was guaranteed for EF traffic only, which means that packet loss could occur for AF traffic in case of failure.
Study results on cost effectiveness
As shown in Figure 3, the FMO2 strategy required 46 percent fewer optical transponders over the 5-year period than the PMO, and even 10 percent fewer than FMO1. FMO2 provided 47 percent greater cost savings than PMO, in the initial years, as well as 51 percent savings over FM01 in Year 1.
These findings reflect the fact that optical transport network costs are chiefly determined by the number of optical transponders required and by wavelength consumption. Transponders are by far the most expensive components in the optical transport path, and their count is closely related to router port requirements. Wavelength consumption can impact the scaling requirements of intermediate ROADM systems that are switching the wavelengths.
Figure 3. Summary of network TCO savings
Results over time
During the initial years, the study’s transport network build-out was driven by 100 Gigabit Ethernet (GE) connectivity requirements for the IP link topology. And many wavelengths were lightly loaded.
FMO1 started out with the largest costs, because its connectivity requirements were higher than those for PMO and FMO2. That is, FMO1 needed 1+1 link redundancy to protect EF and AF traffic, which resulted in a full mesh.
The PMO and FMO2 link topologies, on the other hand, were only partially meshed. That’s because MPLS FRR and GMPLS guaranteed restoration can dynamically create detours around link failures.
As traffic increased, in the study, and wavelengths filled up, the incremental network build out was primarily driven by capacity growth. Then FMO1 cost savings caught up with PMO’s due to the greater efficiency of FMO1.
FMO1 and FMO2 were virtually tied in the number of router ports required, both needing 37 percent fewer 100GE ports than the PMO over the 5-year period. The PMO used more router ports and optical transponders because it relied on intermediate routers to restore traffic on failed segments through MPLS 1+1 protection and FRR. FMO1 and FMO2, on the other hand, could restore optical link segments in the transport layer itself through GMPLS.
The FMO1 and FMO2 strategies were also able to deploy transport layer shortcuts and build direct adjacencies between routers. These capabilities reduced the number of router hops in the data path and, consequently, the number of router ports and optical transponders required in the network.
The FMO2 using multi-layer protection and restoration was more cost efficient than FMO1 in the initial build-out years of the network, and effectively accelerated the cost savings of FMO1 by 4 to 5 years. The reason is that both the FMO2 and PMO link topologies only needed to be partially meshed due to their use of dynamic restoration (MPLS FRR over GMPLS GR), while the FMO1 link topology was fully meshed due to the use of 1+1 link protection.
The FMO1 strategy had a higher connectivity requirement than FMO2 and PMO in the initial years. But its cost effectiveness rapidly caught up in later years, when the further network build-out was driven by incremental capacity needs. FMO1 surpassed the PMO in cost efficiency in Year 1.5, because GMPLS-based protection of aggregate traffic at the optical transport layer is more resource efficient than using MPLS-based protection at the routing layer.
FMO2 was able to maintain its initial cost advantage in later years because it benefited from the same incremental GMPLS cost savings as FMO1. And at the same time, it enjoyed additional savings from deploying GMPLS UNI. Cost savings were predominantly achieved through the various methods used to carry EF and AF traffic in each mode of operation, because best effort traffic was unprotected across all modes of operation, but with use of 1+N passive redundancy.
Is availability maintained when cost-effectiveness improves?
The Nokia Bell Labs study examined the same three network protection and restoration schemes to see whether reducing costs had a negative impact on service availability. To quantify the answer, the study calculated the average service availability for each traffic category, in each mode of operation.
Figure 4. Service availability comparison
As Figure 4 shows, any of the three design options can meet service availability expectations, for a given link availability, failure rate, and mean time to repair of fiber cuts.
However, FMO1 and FMO2 met these levels of service availability far more cost effectively. And the multi-layer protection and restoration used in FMO2 offered the highest and quickest return on investments.
Capitalizing on IP/optical control integration financial white paper