These changes mean that planning core networks based on projected traffic peaks is simply too inefficient and economically irresponsible to be sustainable over the long term. On the other hand, packet networks have proven to be more robust, intelligent and flexible, and better equipped to deal with variations in traffic patterns without compromising service quality. Unfortunately, the equipment has also been slightly more expensive on a per port basis.
Service provider networks today
Today, most service providers have a three-layer network: the optical transport layer, the circuit switching layer, and the IP/MPLS layer. Each of these has unique requirements for redundancy, require independent provisioning and usually have a separate team managing them.
When you "peel open" the circuits in the core of service provider networks, you often will see packets from the applications described above. In addition, this packet traffic growth is rising according to most Tier 1 providers, who estimate that the packet/circuit ratio will increase is from less than 50/50 to 70/30 over the next couple of years.
One way they have been trying to cope is by establishing parallel networks to handle the new data, but this only adds to the operational complexity of having multiple network layers (the lower ones focused exclusively on circuits) in the first place.
To deal with the challenges of rapidly growing and unpredictable traffic, a typical IP routing vendor will likely recommend buying bigger and faster routers since all the traffic is IP anyway. While this solution may provide the flexibility to manage dynamic traffic, it doesn't address the growing economic challenge service providers are facing.
On the other hand, if you're talking to an optical vendor, they will argue that their technology is the cheapest and simplest. But with ever-changing traffic patterns, can they ever hope to keep up?
Obviously both approaches are aimed squarely at selling more of the same as legacy vendors try to solve the problems in ways that protect their market share.
The promise of hybrid MPLS/optical solutions
There is, however, an alternative approach to solving problems in the core network. Instead of adding to one layer or another, why not integrate them? Instead of trying to force-fit one technology to be the panacea, why not take the best of each available approach?
As carriers look at the total cost-of-ownership of MPLS, Optical Transport Network and hybrid networks, the benefits of statistical multiplexing where it is indicated - for most of the traffic coming onto the network - become apparent. Juniper's modelling shows savings ranging from 40-65 percent depending on traffic characteristics such as flow size, peak-to-average ratios, and simultaneous traffic peak occurrences.
These factors impact the decision on whether to keep traffic in the MPLS domain or move it to optical domain. The optimum solution is to combine the benefits of a packet infrastructure with the simplicity of switching and optical transport into a converged supercore where aggregation and multiplexing happen at the packet-level. Integrating the packet and optical transport layers also improves both the capital expenditure (CapEx) and operating expenditure (OpEx) economics of the core network.
Sign up for Computerworld eNewsletters.