In this installment of Network Slicing 101, we’ll examine the basics of network virtualization as applied to network slicing.
When we hear about network slicing, one of the most common attributes associated with the concept is the notion that the slices will need to be enabled on an “end-to-end” basis. However, very little is said about what needs to happen to the network in order to be able make the end-to-end concept tenable.
Network virtualization is foundation that network slicing is built on
There is no shortage of scholarly articles that attempt to articulate exactly what must be done to the various network domains in order to enable end-to-end network slicing. One particularly detailed account can be found in a whitepaper released by the Wireless World Research Forum. A common thread that tends to run through this, and other papers on the subject, is that idea that at each point in the relevant network domains (i.e., Radio Access Network, core, transport) the user plane must be separated from the control plane. This de-coupling of user and control plane also happens to be one of the fundamental underlying tenets in SDN/NFV (i.e., network virtualization). Invoking a maxim of Euclid (“Things that are equal to the same thing are equal to each other”), the following must be true:
- Separation of the user and control planes equal the ability to virtualize network elements
- Separation of the user and control planes equal the ability to create end-to-end network slices
- The ability to virtualize network elements equals the ability to create end-to-end network slices
If so, then why?
If one is looking to split hairs, then an argument can be made that the ability to logically separate networks has existed for a long time. After all, it is possible to partition the same physical network to support different access types and service-level agreements in support of different use cases. However, if one accepts the argument that for network slicing to meet all the demands that will be placed on networks in the 5G era, then the idea that network slicing must, ultimately, be built on fully virtualized networks is relatively straightforward.
To support true network slicing, as opposed to SLA-based network partitioning, the ability to slice the network must be dynamic and practically limitless in scale. In order to meet both of these objectives, the partitioning of monolithic physical networks is not tenable. By contrast, if the fundamental promise of network virtualization is the ability to build networks using cost-effective white-box hardware, and the dynamically control network resources via SDN-based orchestration, then the inexorable link between virtualization and network slicing is clear.
