Page 199 - Big data - Concept and application for telecommunications

Big data - Concept and application for telecommunications 5

Under the framework of bDDN, data from all network ingredients at various levels and different
ramifications could be stored and analysed to produce an overall optimized solution for fault
detection and troubleshooting, which can cut down operating expense (OPEX) and capital
expenditure (CAPEX). More intelligent self-maintaining scripts and algorithms would be invented for
resource allocation and network operation to free engineers from cumbersome manual processes.
Moreover, with inclusive information, data-driven OAM will have a marvellous sense of popular
application trend, which may reveal business opportunities, as well as potential security pitfalls,
which may conceal business catastrophes.
4) Big-data-driven autonomous network control
The big data plane can provide efficient methods for optimized network control compared to other
methods when the network behaviour is complex, dependent on many parameters, and changes
frequently over time. In fact, the big data plane allows the following and understanding of network
behaviour so that control parameters can be autonomously updated during network operation for
achieving optimal performance in any given network condition.
In principle, using real-time performance measurement tools, the big data plane can infer an
association rule between performance, network status, and network control parameters, thus
learning the set of parameters that probabilistically maximize network performance.
5) Application-driven network

In a network plane, it is required that the network can be programmed by the application flexibly.
Dynamic programmability is one of the key technologies which enables the application to program
the network infrastructure (NI) flexibly; this is the concept of application-driven networking (ADN).
See Appendix I for the concept of close-loop bDDN-ADN-NI.
6) Accelerating network architecture evolving
By replacing the network state complexity with big data computing complexity, the burden of
network OAM can be greatly alleviated. In other words, as the network architecture is free from
burden, it can evolve again, in the direction of becoming more and more flexible, extensible and
scalable to match the increasingly challenging application requirements.

10 The high-level capabilities of big-data-driven networking

According to the framework of big-data-driven networking, the bDDN framework is made up of three planes
– big data plane, network plane and management plane. The high-level capabilities of big-data-driven
networking are categorized into three parts:
– high-level capabilities of big data plane;
– high-level capabilities of network plane;
– high-level capabilities of management plane.

10.1 High-level capabilities of big data plane
10.1.1 Network big data sensing layer capabilities

1) Capability of DPI-based network sensing
A network sensing layer is required to collect comprehensive data from different layers of the
network plane and management plane. Since collecting and storing all network traffic data is not
feasible in practice, comprehensive network data collection does not mean to collect all network
traffic data. Big-data-driven networking needs to make use of DPI technology to collect interested
network traffic and applications for the analyser.

2) Capability of big data collection
A large amount of data is hidden in the network, including traffic data; alarm data, invalidation data
and log files from network devices and cloud infrastructures; signalling data, configuration data and
routing data from transaction systems; environmental data from external resources, and so on. The

Network and infrastructure 191