IP routing virtualized network functions (VNFs), such as virtualized routers and mobile gateways, offer a wide range of benefits for network operators and their customers. In this blog, the first of a series, let’s discuss the current status of VNFs and potential barriers to adoption.
BACKGROUND: Carrier-grade VNFs
Software-defined networks (SDN) and network functions virtualization (NFV) have taken the telecom industry by storm. Moving telco infrastructure onto telco cloud server-based platforms promised to deliver many benefits:
- CAPEX and OPEX savings
- Reduced time to market for new services
- Increased service flexibility, agility and scalability
It’s no wonder SDN and NFV have attracted the interest of many network operators and service providers.
Over the last couple of years, the market has seen VNFs evolve from proof-of-concepts to multi-vendor NFV interoperability tests, limited field trials, and some early deployments.
Virtualized network functions include implementations of many network elements previously realized as combinations of dedicated hardware and associated software. Virtualized routers can support a range of IP applications, including:
- Route reflector (RR)
- Provider edge (PE)
- Broadband network gateway (BNG)
- Residential gateway (RGW)
- Application assurance (AA).
Similarly, virtualized mobile gateways can be deployed as:
- 3rd Generation Partnership Project – or 3GPP - System Architecture Evolution gateways (a serving gateway or a packet data network gateway)
- Evolved packet data gateways (ePDG) or
- Gateway GPRS support nodes (GGSN)
VNFs and associated applications can greatly vary in their functional characteristics and deployment requirements. While agility and flexibility may come as “natural” benefits of server-based architectures, there could be great variations in functionality, scalability, and performance among different VNF implementations available on the market today.
The evolution to NFV does not diminish the requirement to support “carrier grade” characteristics. Advanced VNFs able to deliver the same levels of “telco grade” features seen in physical routers will reduce the barrier to adoption and will be among the 1st to find their place in virtualized network infrastructures.
Requirements for advanced IP routing VNFs
When we examine IP routing VNFs in particular, we see a number of key areas that need to be addressed to lower the barriers for introduction into service providers’ networks.
An advanced routing VNF must be a result of sophisticated software design. Optimized for the telco cloud server architecture, the VNF must maximize the use of all computer and networking resources and capabilities.
Control plane (IP/MPLS functions) and data plane (packet forwarding functions) have different requirements for performance, scaling, and resiliency. Often a specific combination of these requirements is required for a particular application. For example, the BGP route reflector application requires intensive control plane processing but asks for a minimal usage of data plane resources. Similarly, a provider edge router or a broadband network gateway application requires performance and scalability for both control plane and data plane functions.
In order to deliver high performance, a virtualized router – a software implementation of a routing VNF - needs to be able to use all available resources (CPUs, CPU cores, memory, I/O resources) efficiently and elastically. Efficiency relates to optimal distribution of processing across resources. Elasticity refers to ability to adjust to actual requirements by adding or removing resources and scheduling tasks accordingly.
These aren’t trivial tasks. That’s why it is challenging even for major (physical) router vendors to implement an advanced IP routing VNF—let alone for NFV startups without a lot of experience and market traction.
An advanced routing VNF must support both horizontal (scaling out – across telco cloud servers) and vertical scaling (scaling up within a VM) for a large number of high-throughput packet flows. In addition, these flows may be using different control planes (Ethernet, MPLS, IPv4, and IPv6). It’s therefore important that both data and control planes are able to provide cloud-scale capabilities.
Moreover, this scalability must support scenarios where multiple IP and MPLS routing services and applications are supported concurrently, which is often the case when deploying routing VNFs for business, mobile, and residential services.
Service providers are accustomed to high resiliency standards when using custom hardware routing platforms and expect same levels of resiliency when deploying routing VNFs. Graceful recovery mechanisms are required at both control plane and data plane. The software design of an advanced routing VNF must also support resiliency capabilities enabled by the underlying server-based infrastructure.
The VNF management capability is very important. It goes beyond the virtual machine management – extending to “proper” router management (configuration, provisioning, and maintenance). This is an area where experienced vendors can leverage their tried and tested principles and approaches used with existing physical routing infrastructure.
As virtualized-only environments are still unrealistic for most service providers (especially large ones), common routing management requirements are also imposed on routing VNFs - in order to facilitate a quick introduction of IP routing VNFs and their integration in hybrid network environments.
IP Routing VNFs: Are they ready for prime-time deployment?
With all of this in mind, the pressing questions are:
- Are IP routing VNFs available today ready for commercial deployment?
- Do they have carrier-grade characteristics?
- Can we rely on the performance, scalability, and resiliency of virtualized routers the same way we can rely on the physical routers to do the job?
We at Nokia think so.
In January 2016, EANTC (European Advanced Networking Test Center) – well recognized for their objectivity and vendor-neutral network performance testing – performed independent testing and validation of Nokia’s advanced virtualized routing and networking functions. This included routing VNF applications using the Nokia Virtualized Service Router (VSR) and virtualized mobile gateway applications, using the Nokia Virtualized Mobile Gateway (VMG).
For details about this testing and results, please visit http://lightreading.com/nokiaNFVtests2016report.
In upcoming blogs, we will dig deeper into the key areas of performance, scalability, resiliency, and manageability. We’ll cover major use cases for routing VNFs, along with deployment considerations that are crucial for a smooth journey into the evolving virtualized world. Stay tuned.
We look forward to your questions and comments.