Most discussions I have these days about network functions virtualization (NFV) include a mention that carriers will demand new revenue-generating services out of the technology. But here at the SDN & OpenFlow World Congress, NTT Docomo spelled out what it wants out of NFV, and it’s all about cost savings and service reliability — not new services, at least not immediately.
At the same time, NTT is expanding its experiments with virtual evolved packet cores (vEPC) for LTE networks, said Tetsuya Nakamura, speaking as part of an Open Networking Foundation (ONF) workshop that kicked off the conference Tuesday.
Here’s what he said Docomo expects to get out of NFV. What looms large is not the need for new services, but the need to keep services running after a natural disaster:
- Flexible traffic control — namely, the ability to allocate bandwidth on demand should call volumes suddenly spike. Let’s face it, everyone’s first instinct after a disaster is to call home.
- High availability — giving the network a way to heal automatically in case of a failure. What this and the first point have in common is that they let Docomo guarantee a service from end to end.
- Capex reduction — and avoiding vendor lock-in, Nakamura implied: “If we can separate the functions on a component basis, maybe we can choose the best kind of component” for each job and create “a more competitive environment,” he said.
When you combine that component approach with the point about end-to-end quality-of-service, though, you realize Docomo will have to make sure that a variety of virtual network functions (VNFs) can be integrated properly.
That’s why Docomo is being meticulous about the vEPC. So far, the carrier has done proofs of concept (PoCs) with three vendors — Alcatel-Lucent, Cisco, and NEC — with each one providing a complete package: virtual EPC; hypervisor; management and orchestration; and the hardware to run the stuff on.
Docomo is still keeping it relatively simple. The combinations consist of one vendor’s vEPC and another vendor’s package of hypervisor; management and orchestration; and hardware, Nakamura said. “All of the possible combinations are being tested with our partners,” he added.
SDNCentral is in Dusseldorf, Germany for the 2014 SDN & OpenFlow World Congress. Check out all our headlines from the show here: SDN & OpenFlow World Congress: The SDNCentral Report