Applications Deployment Flexibility in NFV
Network Functions Virtualization (NFV) has resulted in replacing dedicated network appliances – such as routers and firewalls – with software running on commercial off-the-shelf servers. These software or applications running on servers are called as Virtual Network Functions or VNFs.
Based on a recent survey by IHS Infonetics, more than 75% of the respondents rated “increased operational efficiency” as a strong driver for NFV. Flexibility in deploying VNFs (in terms of location of the servers) is a key factor for increased operational efficiency in NFV networks.
Five options are available for the service providers to host their NFV servers.
- Provider Central Office (CO)
- Provider Data Center (DC)/Private Cloud
- Provider Point of Presence (POP)
- Enterprise Head Office (HO)
- Enterprise Branch Office (BO)
Most of the service providers prefer to deploy VNFs at their CO or POP and deliver the services via a router. Such deployments help them leverage the existing CO or POP facilities to address latency, particularly for those services challenged by high bandwidth, heavy traffic and quick response requirements. Moreover in these cases, the VNFs are able to deliver both services needed by big enterprises (with autonomy and access redundancy) and a more economical service for connecting small branches.
Provider DC/private cloud deployment normally addresses a less demanding requirements of a domestic small and medium business such as a firewall or a VPN, that are not constrained by large bandwidth or high QoE as in the case of a higher-paying enterprise customer. In this case, service chaining can divert traffic to other DCs thereby optimizing the utilization of compute/storage resources in one DC, however at the expense of incurring some delay.
Enterprise HO/BO deployments are mostly driven by customers who are more concerned about the security and high-availability of the services. Even though these are the least preferred deployment options, they guarantee an autonomous behaviour in the case of an access fail.
Distributed Deployment of VNFs
A more pragmatic approach is a distributed deployment model where some of the VNFs are running on servers at the Enterprise HO/BO whereas some others at service provider CO or POP or DC. For example, in the case of an Enterprise vCPE, the data plane functions such as NAT, QoS, etc. are deployed at Enterprise HO/BO (Distributed Data Plane). Meanwhile, the control plane functions such as DNS, DHCP, etc. are deployed at a service provider CO or POP or DC (Centralized Control Plane). Distributed deployment of VNFs brings in the advantages of network offloading capability, reduced latency & jitter considerations, higher reliability & availability and improved security.