It is clear, for Service Providers to support the exponentially growing demand for data bandwidth with declining to flat ARPUs (Average Revenue Per User) their network strategy has to evolve. The heavy capital expenditures of the traditional siloed network functions running on bespoke hardware is a relic of of the past. Market saturation combined with new entrants to the market has created price competition that simply isn’t sustainable with the current network model.
Enter, the web scale companies such as Google, Amazon and Facebook. They have taken a new approach to raw computing horsepower by leveraging low-cost COTS hardware to provide high availability at massive scale instead of the silos of high-cost proprietary appliances. We have seen the adoption of a web scale architecture from companies such as AT&T and their Domain 2.0 initiative which is promising 75% of their network to be virtualized by 2020. Obviously this is a bold move that requires immense planning and careful vendor selection in this wholesale change of the network architecture. Margaret Chiosi of AT&T lays out some of these considerations in this interview:
Margaret does a good job of framing 3 challenges of moving towards Network Function Virtualization, or NFV for short, for large service providers.
1. When a vendor moves from ASIC specific hardware to COTS, how do you get throughput performance?
2. With virtualization, functions can move around, we (AT&T) need full automation since the platform can grow and shrink. The industry does not have a “common framework” for full automation.
3. How do you re-optimize a NFV that has been ported from custom hardware to COTS, is the function architecture optimized? Of course you can use DPDK and SR-IOV but there still can be throughput issues due to the fundamentalarchitecture of an application.
As we have seen, there is no shortage of vendors today that are offering virtualized network functions. But you have to ask yourself how many of these applications were built from the ground up with a virtualized architecture in mind?
Affirmed Networks is ideally positioned to develop virtualized network functions since we don’t have the technical debt of porting our Mobile Content Cloud from a bespoke platform. As Margaret points out, Affirmed Networks does have the deep technical knowledge in its employees regarding high-scale and high- availability network functions.