Finally ready for prime time: multiservice consolidation over MPLS now makes technical and business sense - Network Core

Telecommunications Americas, Nov, 2003 by Riad Hartani

* Efficient QoS in the backbone;

* Efficient restoration schemes;

* Appropriate control driven dynamic provisioning and management schemes.

In addition to the technical aspects, carriers look carefully at the business case for convergence, specifically the timing of migration and selection of services to be migrated. From a purely technical perspective, three questions must be addressed:

1. Are QoS schemes effective enough to emulate services over an IP/MPLS backbone? Specifically, are differentiated services based QoS schemes used in IP/MPLS backbones sufficient for the various emulations? The emulation of some services such as ATM and TDM require very stringent loss, delay and jitter tolerances, and it is still unclear if a differentiated services enabled network is capable of providing these while also satisfying service providers' requirements for higher network utilization. Some form of quantitative guarantee is required to satisfy these requirements, requiring some evolution of existing per-hop behavior schemes used in differentiated services.

2. Are IP/MPLS backbones sufficiently reliable and scalable to carry private and scalable to carry private and business traffic? While IP/MPLS networks are becoming highly reliable and available, work still must be done to get network and network element reliability to the required level, principally regarding the control plane. New specifications based either on graceful restart mechanisms or non-stop routing mechanisms are being deployed now, complemented with new schemes for IP- or MPLS-layer fast restoration and efficient protection.

3. Are IP/MPLS network provisioning, operation and management schemes good enough to satisfy the service delivery and scalability required by multi-services over IP/MPLS? MPLS is being augmented with a number of new features, on the data (e.g., OAM packets), management (e.g., multi-services MIBs) and control (e.g., LDP and BGP for circuit information) planes to facilitate operations. Additional work is being done to ensure that the network infrastructure is ready to handle all the management and operations aspects required of a converged backbone.

The business case for convergence is seductive, promising massively lower capital and operational costs, but two key business questions must be addressed. First, does it always make sense to carry multiple services over IP/MPLS? The answer depends on the network being considered and on the proportion of each traffic type, its growth rate, the selection of the service being emulated, the timing of the migration over MPLS and the cost of the convergence strategy.

Most large carriers have a number of networks: TDM, ATM, frame relay and IP/MPLS, among others. Most are looking at migrating frame relay and ATM traffic directly over IP/MPLS and adding new Ethernetbased services over IP/MPLS with TDM over IP/MPLS planned for the longer term. In other networks, migration is done by capping ATM and frame relay growth and emulating the services over IP/MPLS. In other cases, a converged IP/MPLS backbone may not be the right option at this time. At any rate, convergence over IP/MPLS will be a long process, and strategies will be very service provider-specific.