Page 99 - 5G Basics - Core Network Aspects

Core network aspects 1

8.3 Topology awareness and quick reconfigurability

Since network virtualization allows aggregation of virtual resources distributed in networks, it is necessary to
discover these virtual resources. Furthermore, when constructing or reconfiguring an LINP, optimization may
be required for effective use of the virtual resources. For example, for the users who want low end-to-end
delay, topology awareness may help to provide low delay by using the shortest route, but still provide large
bandwidth route for users who want large bandwidth and irrespective of delay.

Therefore, network virtualization should support topology awareness so that the virtual resources can
effectively interact with each other during the construction of LINPs.
After creating LINPs based on users' requirements, the capabilities of LINPs need to be modified due to
various reasons, for example the changes of users' requirements, the status of networks, policies of the
virtual resources owners, and so on. Hence, each LINP needs to adjust its capability according to the changes
of requirements and the reconfiguration should be quickly done in order to minimize service disruption.
Therefore, the network virtualization should offer methods for easy and rapid creation of LINPs and for
dynamic reconfiguration of them.

8.4 Performance

Network virtualization is typically implemented by introducing a virtualization layer or an adaptation layer
and the virtualization layer creates and manages LINPs. The virtualization layer is a layer between physical
hardware and software running on a physical resource. This layer enables the creation of an isolated partition
of the physical resource. Each partition is designed to accommodate different architectures and applications.
The virtualization layer separates the hardware from the operating system environment. Network
virtualization comes at a cost of reduced performance due to the virtualization layer. Network virtualization
architecture includes an isolation partition with hardware device drivers, I/O stack and applications, placed
over the virtualization layer that supports the control of physical hardware device controllers. This additional
virtualization layer adds overhead and degrades system performance including higher CPU utilization and
lower bandwidth. Thus, the performance of the LINPs may not be as good as the non-virtualized network.
Therefore, the performance degradation should be minimized.

8.5 Programmability

An LINP may be equipped with a programmable control plane and data plane so that users can use
customized protocols, forwarding or routing functions in the LINP. In order to provide flexibility to the LINP,
it is required to implement new control schemes on virtual resources. Programmability can support flexibility
in the control plane and make it possible to easily adopt new control schemes on LINPs, and also in the data
plane to enable various kinds of data processing.
In the control plane aspect, programmability involves control in LINPs such as routing, switching and
monitoring for realizing proprietary control of traffic on an individual LINP. It also addresses parsing new
communication protocols to be utilized in FNs. In the data plane aspect, it refers to data processing inside an
LINP such as transcoding and data-caching for realizing new capabilities for data processing inside the
network to enable innovative network services.
Therefore, network virtualization should support both control and data plane programmability in order to
provide flexibility and evolvability of networks using new control schemes and new data processing
capabilities. Also, each LINP should support the free deployment of control schemes or network architecture
independent of other LINPs or physical networks.

8.6 Management

Each LINP can be a flexible aggregation of physical resources and virtual resources with appropriate network
topology. From this perspective, a number of associations of not only physical-to-physical resources but also
physical-to-virtual resources, and vice versa, have to be managed, which is not common in legacy network
management. These complicated mappings cause difficulty in management, so visibility is required to
understand all interconnections between physical and virtual resources over the physical networks.

89