There has been a lot of attention given to Network Functions Virtualization (NFV) and how it is driving a paradigm shift in the architecture of today’s communications service providers (CSPs) networks. NFV is a popular topic and has been getting a lot if visibility in the recent months. Recently, I have been seeing announcements from vendors stating that they are delivering a software version of their application or solution eliminating the need for proprietary hardware. All of these announcements always mention that since the product is now available as a ‘virtualized’ solution, they are NFV compliant or based on NFV standards.
Let’s get something straight. First, there are no standards. There is an ETSI NFV working group that is defining the requirements for NFV. They have produced an initial white paper describing the benefits, enablers, and challenges for NFV in the CSP environment. Most of the industry is using this white paper as the basis for their concept of NFV. The working group is continuing to meet and determine what NFV consists of with a goal of having the concept fully defined by the end of 2014.
Second, and more importantly, just because your solution can be installed and run on commoditized hardware, it does not mean that it is ‘NFV enabled’. In a recent interview, Margaret Chiosi from AT&T said, ‘Orchestration and management is a key part in realizing NFV’. While running on commoditized hardware is part of the NFV story, it is not the complete story. Virtualization does not achieve anything except reduced capital costs due to the use of common off the shelf (COTS) hardware, and flexibility in the deployment of services since there is not a proprietary hardware dependency.
We, at F5 Networks, believe that the orchestration and management of these virtualized services are a critical aspect to make NFV successful. We have been delivering virtualized versions of our solutions since 2010. Since we have experience delivering virtualized solutions, this means that we understand how important it is to deliver a framework allowing for the integration of these solutions with other key components of the architecture.
Even though services such as DPI, Carrier Grade NAT, PGW, and other Evolved Packet Core (EPC) functions have been virtualized, they are not automatically part of a flexible and dynamic architecture. Orchestration and management are necessary to enable the potential of the virtualized service. As I mention in a previous blog post, orchestration is where NFV can really shine because it allows the CSPs to automate service delivery and create an ecosystem where the infrastructure can react and respond to changing conditions. The infrastructure can monitor itself, determining the health of the services and instigate changes based on policies as defined by the architects and operators.
F5 has been providing the foundation for orchestration services ever since the company’s inception. The concept of load balancing and providing application delivery controller (ADC) services is all about managing connections and sessions to pools of services. This functionality inherently provides a level of security since the ADC is providing proxy services to the applications as well as availability since the ADC is also monitoring the health of the service through extensive application health checks.
As Tom Nolle states in one of his blog posts, he likes the idea of ‘creating a set of abstract service models that are instantiated by linking them to an API’. This sounds like a great template for delivering orchestration services. Orchestration is the fact that these service models are linked to the services via APIs. Service models are defined via operator policies and APIs allow for the bi-directional communication between service components.
The application delivery controller (ADC) is a key component for this orchestration to occur. It is the ADC that has insight into the resources and their availability. It is the ADC that manages the forwarding of sessions to these resources. In the CSP network architecture, especially as we evolve to LTE, this includes much more than the traditional load balancers and content management solutions that sit in front of server farms in data centers that we typically think of when discussing ADCs. The high value content is in the packet data network (Gi and SGi) and the control plane messaging (DNS, Diameter, SIP). With this critical role in the real-time traffic engineering of the services being virtualized along with the visibility into the health and availability of the resources the service can access, it makes sense that the ADC play a pivotal role in the management and orchestration of the virtualized infrastructure.
It takes more than content inspection, load balancing, and subscriber policy control via PCRF/PCEF to enable the full orchestration required for the NFV vision to come to fruition. There needs to be an intelligence that can correlate the data from these services and technologies to determine the health and state of the infrastructure. Based on this information along with policies that have been defined by the operator and programmed into this intelligence, it becomes possible to 1) create a dynamic orchestration ecosystem that can monitor the different services and functions within the EPC, 2) collect data and provide analytics, and 3) proactively and reactively adjust the configuration and availability of resources based on the operator defined policies.
Along with the intelligence, it is necessary to have open APIs that allow for inbound and outbound communications to allow for the sharing of the data collected in addition to being the conduit to deliver policy and configuration changes as determined by the orchestration system. It is critical for the orchestration of a NFV architecture in the EPC to be open and allow for multiple, potentially disparate vendors and technologies working together to create a dynamic environment that provides the flexibility and scalability that NFV is looking to achieve.
As an example to demonstrate this orchestration functionality and how it is able to take advantage of virtualization within the NFV architecture, I am reposting this video of value-added services (VAS) bursting in a virtual EPC (vEPC) environment. In this scenario, the F5 BIG-IP Policy Enforcement Manager (PEM) is identifying and tracking the connections being delivered to the VAS solution, video optimization in this case. Based on the data received, such as number of concurrent connections, BIG-IP PEM is able to signal a virtual orchestration client, such as one of many varieties of virtual machine hypervisors to enable additional servers for the video optimization solution and have them added to the available resources within the traffic steering policy.
This demonstration shows the initial potential of a virtualized infrastructure when one is able to deliver on the promise of the orchestration and management of the entire infrastructure as an ecosystem, not a pool of different technologies and vendor-specific implementations. It is critical for this collaboration and orchestration development to continue if we expect this NFV architecture to be successful. It is important for everyone, the CSPs, vendors, and technologists, to see and understand that NFV is much more than virtualization and the ability to deliver services as software. NFV requires a sound management and orchestration framework to become a proper success.