Page 200 - Big data - Concept and application for telecommunications
P. 200
5 Big data - Concept and application for telecommunications
bDDN is supposed to collect all the information that is related to network business and status, while
at the same time fulfilling the mission of storage, classification and analysation.
10.1.2 Network big data repository capabilities
1) Capability of network big data storage
The network contains a variety of a large amount of data; bDDN needs to provide all kinds of large
data storage capabilities, including a mass-structured data storage capability and an unstructured
data storage capability.
2) Capability of big data interchanging and sharing
Different network management realms are supposed to be capable of big data sharing and
exchanging, so that big data resources can be utilized by different realms or network services.
10.1.3 Network big data analyser layer capabilities
1) Capability of correlation analysis
A big-data-driven network analyser conforms to the traditional big data paradigm: comprehensive
data outweighs sampled data, correlation rather than causality. It is difficult for a network designer
or operator to find and compute the network causality. The correlations learned from network plane
data and management plane data can be utilised to effectively improve network resource allocation,
reduce CAPEX and OPEX, and maximize revenue.
2) Capability of big data real-time computing and analysing
Considering the huge amount and fast changing characteristics of network data, bDDN is required
to provide real-time analysing ability. bDDN would perform comprehensive data fusing and
analysing over the collected information, correlate various influencing factors and network status,
and find out the causality and logistics behind them by using big data technologies.
3) Capability of machine learning and deep learning
The network big data analyser layer is the core component of the big data plane. It will take the
machine learning and deep learning on the huge data collected by big data sensing layer, to achieve
network perception and cognition, including network autonomous optimization, autonomous
adjustment, intelligent fault location and a series of network intelligence goals.
4) Capability of network QoS anomaly detection and root cause tracking
Different applications have different QoS parameter requirements, for example, delay/latency,
jitter, round trip time, etc. Network QoS anomaly means network QoS parameters anomaly. To meet
the complex QoS/QoE requirements of different applications/services, the networks are required to
detect the network QoS anomaly and track the root cause of the anomaly. The bDDN should have
the capability to automatically monitor the network QoS anomalies and track the root causes of the
anomalies. The problems of network performance anomalies are network data, such as, network
traffic data, syslog data and management data, etc. The QoS anomaly network data is from network
anomaly events, such as, network attacks, protocol bugs and link up/down, etc. Based on the
analysis of multilayer dependence and spatial-temporal dependence of network data and network
events, bDDN can reversely track the root causes of network anomalies. The bDDN is able to clarify
the positive correlations and reverse tracking mechanisms of network 'anomaly events – anomaly
data – network anomalies'.
10.1.4 Data intelligence and service layer capabilities
1) Capability of big data visualization
The network data visualization is a visual representation of the insights gained from the network
data analyser. A network data visualization capability may reveal the hidden potential value of
comprehensive network data. Network data visualization exhibits the correlations and implications
of raw network plane and management plane data with images, tables, graphs, charts, diagrams
and maps etc., so that network operators can see and understand the connections in a real network.
192 Network and infrastructure