Recently there has been several announcements around innovation for core networks, both from ourselves at Juniper for our T-Series Platform and others; Juniper and Verizon also recently announced the deployment of the PTX in its Global MPLS backbone for 100Gb Ethernet.
While much of the focus of the messaging has been about how we cope with the need to scale the network core - the other critical point that the core network also needs to be simplified may have been overlooked slightly.
We live in a world where massive grow in bandwidth demands in all parts of the network have become the norm; our industry colleagues at Cisco recently published figures which suggested that worldwide requirements for bandwidth would grow fourfold, meaning global IP traffic will reach 1.3 zettabytes per year by 2016. These projections are based upon an aggregation of data from more than 20 of the major analyst firms.
Core networks are about moving packets on a massive scale, connecting users to each other, users to content and delivering advanced services from the edge-to-edge over the network. The Core has a number of fundamental aspects:
Transport – moving packets efficiently between the users at the network edge and connecting them to content and services which are centralized. This transport of traffic is local, national or often global.
Scale – the ability to flexibly meet demands placed upon it by users and service created at the network edge; traffic patterns change over time and the core has to cope with this.
Service Quality – As a key component in the delivery chain Core networks play an important role in delivering a quality of service to a user or service – often against a guaranteed SLA.
Resilience – Core networks have to engineered in such as way as they continue to deliver service irrespective of the failure scenario – this can be achieved with static protection (such as is often delivered by the optical layer), or dynamic restoration (such as can be delivered by MPLS fast reroute).
Cost Effective – in order to provide service providers the ability to deliver services profitably the core has to achieve all of the above attributes at a decreasingly lower cost-per-bit each year.
Green – The largest Core routers are hardly environmentally friendly consuming many kW of power – if you can cut this by more than 50% you can save a lot of power costs and be more environmentally friendly.
To achieve all of the above it doesn’t require a more complex Core network it actually requires one which is simplified.
Looking at the layers in the network today; an underlying optical “transport” infrastructure designed to move traffic over and light up the dark fibers in the ground and a packet switched Core Network layer often built up of IP routers. In some legacy networks there is also a circuit switched technology layer which is used to deliver point-to-point circuits (such as E1, STM-1). All of these layers operate independently, adding complexity and therefore cost – collapsing them together, or removing them altogether in the case of the circuit switched layer, has the benefit of decreasing complexity and increasing the efficiency of the Core network. A good example here is the removal of all of the interfaces which interconnect all of these layers – saving the Operator significantly in both CAPEX and OPEX.
The traffic mix in the core network is also changing; According to a recent study by ACG Research , the ratio of time-division multiplexing (TDM) to IP traffic will dramatically change over the next five years. Whereas today, TDM-encapsulated traffic still represents 50-70% of all traffic carried on the core transport networks, this will shrink to 10% by 2016. IP traffic is trending the opposite way, growing from 30-50% to more than 90% in the same time period. In fact, estimates show that, even today, the dominant traffic within (carried over) TDM circuits is packet-based.
Consider the routers in the packet layer of the Core network – they simply need to transport (or forward) packets, which have been labeled at the network edge as quickly and efficiently as possible; thus core routers don’t really need to know about, or hold, the whole IP routing table of the internet, releasing them from this burden – means we can make the next generation of Core routers more scalable, more dense in terms of interfaces per rack, and more cost efficient – not to mention much greener in terms of their power consumption. The industry has settled on MPLS as the defacto standard for building these simplified networks.
Core transport networks are at a crossroads. Service providers are under intense pressure to grow their core networks while reducing costs. Juniper’s Converged Supercore architecture introduces a new network model that enables service providers to take uncertainty and cost out of core network planning and provisioning with an architecture that makes the best use of available technology — creating a core packet transport system that employs MPLS for switching and integration with DWDM/OTN for transport. Junipers vision for the evolution of the core network is based upon the principle of applying most optimised product platform to each network function. In the Core this means simplified: scale, efficiency, ease of operation and flexibility.
– END –