Common Use Cases for the Transition to NFV
By Gary Hamilton
In my previous blog posting I discussed the question – What is NFV? This blog post looks at the network functions that will be delivered as virtual network functions (VNFs), instead of as hardware appliances. SDx Central offers a nice article that helps with some of these definitions.
In very simple terms (remember this is a blog series for IT people, not network experts), a network function is a network capability that provides application and service support, and can be delivered as a composite whole, like an application. SDx Central does a good job in the aforementioned article of grouping these network functions by categorizing them into umbrella or macro use cases. By leveraging these macro use cases, I am able to provide a layman’s description of what these use cases are attempting to achieve, and the service support they deliver. The macro use cases I will focus on in my explanation are “virtual customer edge” and “virtual core and aggregation”, because these are the two use cases that are generally being tackled first from a NFV perspective.
Use Case – Connecting a remote office (using vCPE)
In layman terms, the SDx Central “customer edge” use case focuses on how to connect a remote office, or remote branch, to a central data centre network, and extending the central data centre’s network services into that remote office. In order to deliver this connectivity, a CPE (customer premises equipment) device is used. Generally, the types of device used would be a router, switch, gateway, firewall, etc., (these are all CPEs) providing anything from Layer 2 QoS (quality of service) services to Layer 7 intrusion detection. The Layer 2 and Layer 7 references are from the OSI Model. vCPE (virtual customer premises equipment) is the virtual CPE, delivered using the NFV paradigm.
The following diagram was taken from the ETSI Use Case document (GS NFV 001 v1.1.1 2013-10), referring to the vCPE device as vE-CPE. (I’ll discuss the significance of ETSI in a future blog post)
This diagram illustrates how vCPEs are used to connect the remote branch offices to the central data centre. It also illustrates that it is OK to mix non-virtualised CPEs with vCPEs in an infrastructure. Just as in the enterprise IT cloud world, the data and applications leveraging the virtual services are not aware – and do not care – whether these services are virtual or physical. The only thing that matters is whether the non-functional requirements (NFRs) of the application are effectively met. Those NFRs include requirements like performance and availability.
This particular use case has two forms, or variants –
- Remote vCPE (or customer-premise deployed)
- Centralised vCPE (deployed within the data centre)
The diagram below shows examples of both variants, where vCPEs are deployed in branch offices, as well as centrally. The nature of the application being supported, and its NFRs, would normally dictate placement requirements. A satellite/cable TV set-top box is a consumer example of a “customer-premise deployed” CPE.
Use Case – Virtualising the mobile core network (using vIMS)
The SDx Central “Virtual core and aggregation” use cases are focused on the mobile core network (Evolved Packet Core – EPC) and IP Multimedia Subsystem (IMS). In layman terms, this is about the transportation of packets across a mobile operator’s network. This is focused on mobile telephony.
IMS is an architectural network framework for the delivery of telecommunications services using IP (internet protocol). When IMS was conceived in the 1990s by the 3rd Generation Partnership Project (3GPP), it was intended to provide an easy way for the worldwide deployment of telecoms networks that would interface with the existing public switched telephone network (PSTN), thereby providing flexibility, expandability, and the easy on-boarding of new services from any vendor. It was also hoped that IMS would provide a standard for the delivery of voice and multimedia services. This vision has fallen short in reality.
IMS is a standalone system, designed to act as a service layer for applications. Inherent in its design, IMS provides an abstraction layer between the application and the underlining transport layer, as shown in the following diagram of the 3GPP/TISPAN IMS architecture overview.
An example of an application based on IMS is VoLTE, which stands for “Voice over 4G LTE” wireless network. Examples of VoLTE applications are Skype and Apple’s FaceTime.
Use Case – Virtualising the mobile core network (using vEPC)
While IMS is about supporting applications by providing application server functions, like session management and media control, EPC is about the core network, transporting voice, data and SMS as packets.
EPC (Evolved Packet Core) is another initiative from 3GPP for the evolution of the core network architecture for LTE (Long-Term Evolution – 4G). The 3GPP website provides a very good explanation of its evolution, and description of LTE here.
In summary, EPC is a packet-only network for data, voice and SMS, using IP. The following diagram shows the evolution of the core network, and the supporting services.
- GSM (2G) relied on circuit-switching networks (the aforementioned PSTN)
- GPRS and UMTS (3G) are based on a dual-domain network concept, where:
- Voice and SMS still utilise a circuit-switching network
- But data uses a packet-switched network
- EPS (4G) is fully dependent on a packet-switching network, using IP.
Within an EPS service, EPC provides the gateway services and user management functions as shown in the following diagram. In this simple architecture:
- A mobile phone or tablet (user equipment – UE) is connected to an EPC over an LTE network (a radio network) via an eNodeB (a radio tower base station).
- A Serving GW transports IP data traffic (the user plane) between the UE and external network.
- The PDN GW is the interface between the EPC and external network, for example, the Internet and/or an IMS network, and allocates IP addresses to the UEs. PDN stands for Public Data Network.
- In a VoLTE architecture, a PCRF (Policy and Charging Rule Function) component works with the PDN GW, providing real-time authorisation of users, and setting up the sessions in the IMS network.
- The HSS (Home Subscriber Server) is a database with user-related and subscriber-related information. It supports user authentication and authorisation, as well as call and session setup.
- The MME (Mobility Management Entity) is responsible for mobility and security management on the control plane. It is also responsible for the tracking of the UE in idle-mode.
In summary, EPC, IMS and CPE are all network functions that deliver key capabilities that we take for granted in our world today. EPC and IMS support the mobile services that have become a part of our daily lives, and frankly, we probably would not know what to do without them. CPE supports the network interconnectivity that is a part of the modern business world. These are all delivered using very specialised hardware appliances. The NFV movement is focused on delivering these services using software running on virtual machines, running on a cloud, instead of using hardware appliances.
There are huge benefits to this movement.
- It will be far less expensive to utilise shared, commodity infrastructure for all services, versus expensive, specialised appliances that cannot be shared.
- Operational costs are far less expensive because the skills to support the infrastructure are readily available in the market.
- It costs far less to bring a new service on-board, because it entails deploying some software in VMs, versus the acquisition and deployment of specialised hardware appliances.
- It costs far less to fail. If a new service does not attract the expected clientele, the R&D and deployment costs of that service will be far less with NFV than in the traditional model.
- It will be significantly faster to bring new services to market. Writing and testing new software is much faster than building and testing new hardware.
- The costs of the new virtual network functions (VNFs) will be less because the entry point is far lower because it is now about developing software versus building a hardware appliance. We see evidence of this, where a lot of new players have come into the Network Equipment Provider (NEP) market (the suppliers of the VNFs), therefore creating more competition, which drives down prices.
All sounds great. But, we have to be honest, there are serious questions to be answered/addressed –
- Can the VNFs deliver the same level of service as the hardware appliances?
- Can the Telco operators successfully transform their current operating models to support this new NFV paradigm?
- Can a cloud meet the non-functional requirements (NFRs) of the VNFs?
- Are the tools within the cloud fit for purpose for Telco grade workloads and services?
- Are there enough standards to support the NFV movement?
All great questions that I will try to answer in future blogs. The European Telecommunications Standard Institute (ETSI), an independent, not-for-profit organisation that develops standards via consensus of their members, has been working on the answers to some of these questions. Others are being addressed by cloud vendors, like VMware.
Gary Hamilton is a Senior Cloud Management Solutions Architect at VMware and has worked in various IT industry roles since 1985, including support, services and solution architecture; spanning hardware, networking and software. Additionally, Gary is ITIL Service Manager certified and a published author. Before joining VMware, he worked for IBM for over 15 years, spending most of his time in the service management arena, with the last five years being fully immersed in cloud technology. He has designed cloud solutions across Europe, the Middle East and the US, and has led the implementation of first of a kind (FOAK) solutions. Follow Gary on Twitter @hamilgar.