Network functions virtualization (NFV) can help service providers reduce equipment and operational costs, power consumption, and time-to-market for new services and functionality. By many accounts, NFV will dramatically change the networking landscape. It is leading the industry’s network transformation to a virtualized, cloud-based approach, characterized by:
- An independent generic hardware layer
- Automated network operations
- An agile application development and deployment model
NFV’s software-driven approach lets service providers automate more operations. It can also simplify operational processes, such as capacity planning. NFV stakeholders have formed a specification group within the European Telecommunications Standards Institute. It has identified a common NFV architecture model with an infrastructure layer with compute and networking resources, along with a common management and orchestration layer. These two components comprise the NFV platform on which virtualized network functions can run.
Reliability and efficiency Service providers won’t accept network functions being unavailable if a cloud data center or an entire region loses service. Virtual network functions (VNFs) must fulfill the same strict latency requirements as traditional network functions. They need to:
- Provide five-nines availability
- Offer the same quality of service and security levels as telecommunications networks
Traditionally, network functions have been delivered on purpose-built, telecom-specific hardware. This hardware provides the reliability and long mean times between failures that let service providers achieve acceptable availability levels. NFV uses cost-effective commodity servers and hardware elements that may not be as reliable. Availability is therefore best measured at the system level, where many cooperating elements work in conjunction with the NFV platform. NFV gives service providers a new level of operational efficiency. NFV’s software-based approach enables a much higher degree of automation. Everyday operational processes can be programmed and executed with little or no human intervention (zero-touch). This reduces process intervals from months and years to days and weeks. In addition, operations staff can deliver, install, or repair network elements remotely, without leaving the office.
What should service providers and application developers look for in an NFV platform? It needs to:
- Take full advantage of virtualization (Figure 1)
- Provide a management and orchestration layer, including a VNF lifecycle management capability that simplifies and accelerates deployment and management of VNFs
- Provide an NFV infrastructure consisting of cloud nodes with compute and storage resources, as well as a software-defined networking (SDN) framework to rapidly provision the virtual network structures needed for NFV applications
Together, these elements provide operational tools that will give service providers new horizontal and highly efficient operational models.
Reaping the benefits
To take full advantage of network functions virtualization, the NFV platform must combine virtualization and cloud technologies with the following specifications: 1. Distributed cloud infrastructure To support the distributed NFV architecture, an NFV platform must:
- Give application developers fine-grained control over the placement of network functions and sub-functions
- Automatically find the optimal workload locations based on service provider policies and resource availability
In addition, service providers need to manage and orchestrate the distributed datacenters and networks as a single virtual cloud. This allows operational staff to analyze and monitor the entire cloud platform in real time using tools to correlate events and metrics from various inputs for more effective decision making. 2. Automated cloud nodes NFV’s distributed infrastructure means that datacenters and points of presence must be added and changed more frequently than in a centralized cloud with centralized datacenters. There may be up to thousands of cloud nodes across the coverage area, each of them providing compute, storage, and network resources. To attain the operational expenditure advantages of NFV, service providers need to deploy cost-effective, mass-produced hardware. With this approach, individual components are no longer repaired, but instead are taken out of service until the cloud node is replaced. To keep operational costs low, this requires a highly automated and pre-configured type of cloud node. While it can take a week or more to install a regular cloud node, an NFV cloud node can be bootstrapped and configured in a matter of hours, with most of the work performed remotely. 3. Automated application lifecycle management The pure software nature of NFV applications enables more efficient lifecycle management processes. In the best case scenario, a new service can be deployed at required locations with the push of a button, without having to procure and install new equipment. This entails describing the application components and specifying how to execute the lifecycle phases, including:
- Phase out
Service provider applications often consist of multiple virtual machines, storage volumes, and network configurations. That means service providers need to see how these components group to form applications, what applications are running where, and how many resources they use in addition to seeing the individual virtual machines. To accomplish this, the cloud management system must see the application as a set of VNFs. For example, an IMS application may consist of 10 or more different types of VNFs, each with its own deployment, scaling, and maintenance rules. Today, deploying network elements for a new service and configuring the required network connectivity are separate processes executed by different teams. With NFV, these processes are further complicated by the large number of datacenters and points of presence. And by the frequent changes required when locations are added or removed. This makes traditional manual deployment processes virtually impossible. The ideal NFV platform makes deploying a network application and allocating and configuring the necessary compute, storage, and network resources an integrated and automated process that takes minutes instead of weeks. The process considers application policies and requirements described in the application descriptors. 4. Network automation NFV applications must be embedded in the network at the LAN and WAN levels to achieve the desired performance and availability. To accomplish this, NFV applications such as virtual IMS and virtual EPC often need sophisticated network structures that handle media, signaling, and management traffic differently. VNFs are also much more dynamic than their physical counterparts. They are deployed and removed on demand and scaled to match changing traffic. In addition, communications paths between blades in a physical network function are mapped to communications paths between virtual machines in NFV. These communications paths must be provided by the NFV infrastructure. Achieving this requires rapidly configurable and flexible network abstractions, which classical router and switch networks cannot easily provide. The NFV platform should have access to SDNs, which provide enhanced programmability. Combined, NFV and SDN can capture the network needs of VNFs and create the network capabilities automatically, as needed (Figure 2). This helps service providers resolve the labor-intensive and error-prone configuration processes in today’s networks.
5.An open and multivendor environment The NFV platform must be an open, shared environment capable of running applications from different vendors. Vendor-specific NFV silos with vendor-specific hardware and platform capabilities defeat the purpose of NFV. Service providers must be free to make their own hardware selection decisions, change hardware vendors, and deal with heterogeneous hardware. The NFV platform should also shield VNFs from the specifics of the underlying infrastructure. At the application level, service providers want the freedom to choose best-of-breed applications. NFV platforms must, therefore, support industry-standard application programming interfaces (APIs) for NFV applications to access. This article is excerpted from the Alcatel-Lucent strategic white paper entitled “Why service providers need an NFV platform”. To contact the author or request additional information, please send an email to firstname.lastname@example.org.