EVOLUTION - ANDREW LORD
The evolution of core transport networks
Effective management of the phenomenal growth in network traffic hitting the core network requires evolution of data plane and control plane technologies, which can dynamically handle the multiple terabits of data coming down the line. By Andrew Lord.
other large bandwidth users, is leading to new requirements for a Control Plane capable of real time bandwidth management of large scale resources. There are currently several efforts to develop software architectures, under the collective term SDN (Software Defined Networking) that provide these new features. Possibilities include the prospect of multi-layer networks which jointly optimise multiple network layers from optical to IP, and even scope for network virtualisation, in which optical resources can be assigned directly to clients for their own control. One significant issue to be addressed here relates to how future SDN based network management can interoperate with the existing, legacy OSS/ BSS systems. In summary, the exponential growth in traffic load onto the core network will be met by new data plane and control plane technologies, capable of efficient and dynamic handling of multi-Tb/s, multiple fibre networks in the future.
will together serve to increase the efficiency of optical fibre spectrum by more than 100% depending on the geographical scale of a given network. At British Telecom, we expect to see transponders evolve towards banks of resource units, which have recently been termed S-BVTs or Sliceable Bit Rate Variable Transponders – where individual units will include multiple modulators, and therefore be capable of creating a wide range of multiple carrier superchannels, which will ultimately be “sliced” by the network (using components such as flexgrid enabled ROADMs) and forwarded towards different destinations. Service Providers will be able to plug in additional transponder resources as required, allowing the node to automatically and dynamically provision them as required. This increased level of dynamicity for core networks, coupled with increased need for integration with data centres and
use of Content Distribution Networks (CDN) to distribute popular content more closely to customers and avoiding the core network entirely. Despite these strategies, the continued growth in demand is still putting pressure on the core network. This is being addressed currently by evolution from 10Gb/s to 100Gb/s optical line rates, with scope in the short term to extend to 200Gb/s and 400Gb/s using a combination of higher order QAM and multiple carriers. Both of these approaches are useful but limited: higher order QAM has increased reach limitation (though geographically small countries such as the UK can support high QAM rates such as 16QAM and above) and multiple carriers don’t address the issue of spectral efficiency, implying that the optical fibre continues to fill up. In the short to medium term, these current, coherent- based transport technologies are sufficient to handle even optimistic exponential growths. In the long run the industry can expect to see links between major centres in national networks that require multiple optical fibres, carrying complex optical signals with multiple carriers – known as superchannels. The optical fibre spectrum will need to become more flexible, and able to be divided into dynamically changing, large spectral divisions – known as flexgrid. Meanwhile the transponders will have to be rate-flexible, either by dynamic adjustment of their QAM modulation or by direct change to the electronic modulation rate – known as flexrate. These techniques
ANDREW LORD Head of Optical Research at BT N etwork traffic growth continues to be exponential and in the UK, for example, consumer internet growth has been greater than one hundredfold over the past ten years, driven mainly by the demand for video and also the evolution of access speeds. Access networks provide dynamic flexibility to support bandwidth on demand, often expressed as the difference between the so-called Peak Information Rate (PIR – the rate limited by the specific access technology which could be many 10s Mb/s or even 1Gb/s) and the Committed Information Rate (CIR – the bandwidth available from the network, should all users require it simultaneously). Typically, CIRs are dimensioned to be lower than the PIR, allowing for network statistical multiplexing and utilising the fact that users consume bandwidth at different times. This PIR/CIR differential is very significant for the core network, whose bandwidth consequently is the summation of user CIRs and not PIRs, reducing the requirements for ultra-high capacities in the core, on a scale that might otherwise require many parallel optical fibres. The statistical multiplexing benefit of the IP layer is supplemented by other strategies for maintaining reduced optical core bandwidths – such as the
Actual
Model
3500
3000
2500
2000
1500
1000
500
Bandwidth growth at a large UK core router over a 10-year period in Gb/s (model shows exponential fit). 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
ISSUE 6 | Q1 2016 30
Made with FlippingBook Annual report