Page 92 - Proceedings of the 2018 ITU Kaleidoscope

‎ 2018 ITU Kaleidoscope Academic Conference‎

- Mobility: 5G network is introducing more flexible
3.1 Planning and design multiple-tier mobility management states that can be
Planning and design of end-to-end network slices with the tailored to the nature of devices and services
objective of satisfying diverse and distinct service demanded by vertical industries. For example, some
requirements, for example ultra-low latency for URLLC immobile user equipment, such as mMTC sensor
and very high and stable throughput for eMBB services, devices, are not frequently probed by the network for
requires processing lots of information coming from the their stringent energy efficiency and allowed to update
items listed below. location only when they enter a specific region.
- User requirements
- Service requirements 3.4 Monitoring
- Operation environment (e.g., energy consumption) Monitoring is integral part of operation and management. It
- Business goals performs the following tasks.
- Acquisition of system logs and performance metrics
3.2 Construction and deployment - Analysis of workload (traffic) by using performance
models and metrics
After the planning and design, the network slice - Classification of resource utilization status (e.g., high,
construction and deployment phase comes into effect. In moderate, and low)
this phase, the VNO determines the optimal network
topology and amount of computing and networking
resources for each node and link so that the given service 3.5 Fault detection
requirements are perfectly met. The VNO sends request to Fault detection is also an integral part of operation and
the infrastructure provider to obtain the necessary amount management runs continuously in the system to carry out
of resources. the following tasks:
In the deployment phase, the VNO places virtual network - Analysis of syslog, classification into normal and
functions (VNF) on appropriate nodes that have sufficient faulty stage
resources required by the functions. This process is also - Detection of usual/unusual behavior of users and
called service function chaining (SFC), which is traffic
orchestrated from a centralized controller. - Localization of fault
- Measures for recovery
3.3 Operation and management
Operation and management is the most human resource 3.6 Security
demanding task in the networking business. Automation of Security is a very important issue in operation and
operation, control and management helps in cutting down management. The security system performs the following
the cost significantly. The monitoring, fault detection and
security functions provide input information required by tasks:
the operation, control and management tasks. The - Traffic analysis, deep packet inspection
following functions are performed for the operation and - Identification of security threats
management of network slices. - Isolation of the infected component
- Dynamic resource allocation: The elastic virtualized To realize the automation of above functions, there are
computing and network resources (e.g., optical challenges of ever-increasingly complicated configuration
wavelength and radio frequency/time slots) are of network slices on-demand and the service provisioning
allocated on demand based on the number of active through the interaction within the network system itself and
users or service demand. with outside operating environment. The volume of
- Resource adjustment: The elastic resources are performance measurement data produced by such complex
adjusted dynamically on the basis of the values of networks would be too large and complex to process
their current utilization and corresponding impact in manually by human operators.
the network performance.
- Policy adaptation: The policies for the allocation or 4. MACHINE LEARNING IN NETWORKS
arbitration of limited resource between different types Machine learning techniques are helpful for addressing the
of network slices are dynamically adopted on the basis challenges of achieving automation in the network setup,
of the severity of performance degradation or impact control and management. They provide useful analytics to
in business in case of insufficient resource allocation. extract valuable information from raw data and generate
For example, if limited resources are shared between insightful advices and predictions. Machine learning
emergency services and entertainment services, the enables machines to improve performance, make decision
resource arbitration is executed with higher priority with reasoning, creating and exploiting knowledge. It
given to the emergency services when their demand makes predictions and provides suggestions on the basis of
increases. the results obtained by processing the data sets that are too

– 76 –