Page 198 - Big data - Concept and application for telecommunications

5 Big data - Concept and application for telecommunications

information from all network dimensions and the external environment is critical to help reveal the true face
of network circumstances and problems, distinct from current face-covering tiresome troubleshooting and
misty defence against hackers. The collection data is stored in a data storage repository which is based on
cloud architecture.

Though the data contains all the necessary information, to genuinely solve network problems and decrease
network complexity, a data computing and analytics layer is needed, where assorted network models and
cloud computing ability is applied. The computing methods, models, abilities and even the form of results
can be either preconfigured or freshly issued by the data intelligence layer. In spite of these analytical results,
service is needed to make the answers more satisfactory. No matter how intelligent, the raw result is still
quite neutral and takes on a primitive form of metadata. The data intelligence and service layer need to
further process the result into handy information and clear instructions (according to the topology, device
manufacture, network status and so forth) and provide end-to-end network intelligence to the other two
planes in the form of a service.

The network plane is composed of three layers. They are "network infrastructure layer", "network controller
layer" and "network application layer". The network infrastructure layer includes all kinds of network devices
performing packet transmission. The network controller layer includes network controllers, responsible for
policy making and dispatching. The application layer includes different kinds of user applications, such as
network reconstruct, security, QoS, load balance, etc.

The management plane is also composed of three layers: "network operation", "network administration" and
"network maintenance". Taking advantage of data analysis and end-to-end network intelligence in the big
data plane, the management plane realizes a series of network automations, including smart maintenance,
troubleshooting, configuration and optimization. It can further bring about elaborate fine-grain network
operation according to user requirements and feedback, which can provide customers with a better network
surfing experience.

There are two autonomous circles in the framework: autonomous control circle and autonomous
management circle. The autonomous control circle is formed between the big data plane and network plane,
while the autonomous management plane is between the big data plane and management plane.
The bDDN model possesses the following features:

1) Introducing big data plane to enable end-to-end network intelligence
With the help of virtualization and an ever-increasing handling capacity, the computing and storage
ability is practically unlimited. Most importantly, the design of computing and storage in bDDN is
supposed to be planned together with other network infrastructures. Under the architecture of
bDDN, all network elements can become not only visible but also operable. Such end-to-end
visibility is critical in reshaping network intelligence by inventing dynamic models in real-time
fashion. Network parameters would also be mapped live to network operation, with network
resources dynamically allocated.
2) Introducing machine learning to future network architecture
With the development of a network, it is necessary to introduce big data technology and machine
learning to achieve autonomous management and adjustment of the network through the
collection of large amounts of data on the network state. The areas of machine learning which are
easier to be used in the network field may include: troubleshooting of network problems, network
traffic prediction, traffic optimization adjustment, network security auditing, etc., to implement
network perception and cognition.
3) Big-data-driven autonomous network management
The biggest advantage of bDDN is that, such simple network infrastructure plus highly sophisticated
data computing and analysis can result in big-data-driven autonomous network operation,
administration and maintenance (OAM) – data-driven network operation, administration and
maintenance.

190 Network and infrastructure