Page 201 - Big data - Concept and application for telecommunications
P. 201

Big data - Concept and application for telecommunications                       5


            2)      Capability of end-to-end network intelligence provision

                    Based  on  the  management  and  analyzation  of  assorted  network  data,  end-to-end  network
                    intelligence can be provided to network management, QoS, security, maintenance and operation;
                    this  makes  all  these  aspects  more  smart  and  proactive  through  big  data  technologies.  Data
                    intelligent services can be provided both on the management plane and network plane.

            10.2    High-level capabilities of network plane

            10.2.1  Application layer capabilities
            The  bDDN  network  application  layer  is  where  applications  specify  network  services  in  a  programmatic
            manner.  The  network  application  layer  interacts  with  the  network  control  layer  via  application-control
            interfaces. The network application layer needs to interact with the big data plane.

            In the network application layer, it is required that the network can be programmed by the application
            flexibly. Dynamic programmability is one of the key technologies to enable the application to program the
            network flexibly.

            10.2.2  Network controller layer capabilities
            The bDDN network controller layer provides a means to control the behaviour of the network resources layer.
            Instructed by the bDDN network application layer, the bDDN controller layer interacts and instructs the bDDN
            resource layer to transport and process network traffic. The network controller layer needs to interact with
            the big data plane.

            The controller in bDDN needs to communicate with the big data plane, the calculation of policy, forwarding
            rule and so on is computed by the big data plane, and then dispatched to the infrastructure layer by the
            controller. This work in the SDN is completed by the controller with a simple calculation.
            On the other hand, the policy computation in bDDN is based on the large collected data sets, including some
            external  data;  this  is  not  in  the  SDN.  In  addition,  bDDN  also  uses  machine  learning  to  optimize  the
            computation model. Therefore, policies and rules through bDDN computation are more optimized than those
            provided by the SDN controller. The bDDN can make better use of network resources, and provide a better
            network service for users.

            10.2.3  Network infrastructure layer capabilities
            The bDDN network resource layer is where the network elements perform the transport and the processing
            of data packets according to the decisions made by the SDN control layer. The network infrastructure layer
            needs to interact with the big data plane.

            10.3    High-level capabilities of management plane

            10.3.1  Capabilities of network operation layer
            Based on the big data from the network, operators can deeply analyse the real-time status of the network,
            accurately predict the future status of the network, and effectively configure the network to support better
            services.
            10.3.2  Capabilities of network administration aspect

            By analysing the network big data in real time, bDDN can get instant knowledge of network status, existing
            problems  and  undergoing  risks.  Through  coordinating  with  the  controllers  of  the  network  plane  and
            management plane, bDDN can fulfil autonomous network control and management. Accordingly, operational
            expenditure and capital expenditure would be both cut down.
            10.3.3  Capabilities of network maintenance aspect

            Rather  than  frequent  intervention  whenever  a  problem  occurs,  to  circumvent  observable  latency,
            optimization  must  be  done  beforehand.  The  pattern  would  transform  from  "observe,  catch,  diagnose,
            bailout"  to  "overall  circumstance  awareness,  trend  prediction,  pre-emptive  resolution".  The  more  data


                                                                           Network and infrastructure    193
   196   197   198   199   200   201   202   203   204   205   206