OpenStack users share NFV experiences and recommendations for the Mitaka release cycle.


Network Functions Virtualization (NFV) is redefining the way data is distributed by creating highly automated and more efficiently structured networks that deliver next-generation apps and services with ease and speed.

Superuser reached out to Toby Ford, assistant vice president of cloud infrastructure and platform architecture and strategy for AT&T, Anthony Veiga, a network engineer at Comcast, Jincheol Kim, principal research scientist at SK Telecom and Kyle Mestery, former project team lead (PTL) for the Neutron project, for their take on the key benefits of NFV, what should be accomplished in the Mitaka software release cycle and the state of NFV at their organizations.

Several users also shared how their organizations are working with NFV on OpenStack in a report released today by the OpenStack Foundation.

What are the benefits of NFV?

Ford: NFV drastically reduces cycle time in creating and dropping services and/or infrastructure applications. It’s also a more efficient way of delivering on-demand services adjusted to usage.

Kim: First of all, the management of network infrastructure will be enhanced in a more advanced way that can greatly reduce faults and service downtime. Automating what was previously manual network management also reduces faults coming from human errors. This will bring down the cost of network operation and management and provide more effective network resource utilization.

Second, we think NFV will be one of the foundation infrastructures for SK Telecom’s transformation to an advanced platform business company with next generation telecom infrastructure. NFV will turn our telecom infrastructure into an agile, flexible, and programmable infrastructure that can create more business opportunities for our B2B business partners. This will make the Korean ICT business ecosystem with SK Telecom’s telecom services more rich, diverse and dynamic with new business opportunities.

Third, we also expect that our core telecom services can be evolved in a more diverse way and give our company various technological choices for enhancement. This includes outsourcing part of our telecom services using cloud-based functions such as the recent ClearWater project. This kind of possibility can be an opportunity or a threat to our telecom business model, but we think that this kind of recent trend can help SK Telecom focus on advanced telecom services offering significant value-add for our customers.

Mestery: Benefits include the ability to dynamically provision, deploy and remove network services. Elasticity of services is very important and a key driving benefit of NFV. These services can be physical or virtual, but either way, the provisioning aspects are the main benefit.

Veiga: NFV has potential applications across our organization, including outside my area of focus. In my area, which is our OpenStack-powered Elastic Cloud, NFV has the potential to drive a number of benefits. Since NFV moves functions from a bare-metal appliance model to a software-based orchestration model, deployment times will become faster. It becomes possible to expose some control of functions directly to customers, as they can now have dedicated instances. Physical inventory management activities are reduced as they simply become maintenance of available compute capacity in the cloud.

What are you hoping is accomplished in the Mitaka software release regarding NFV?

Ford: The community should focus on Neutron enhancements that support the different SDN controllers and L3 services we’re looking to implement NFV on. We’d also like to see sophisticated scheduling that handles more than initial placement, something akin to VMware’s Distributed Resource Scheduler (DRS).

Mestery: If the Neutron team can come together and finally address “trunk ports to VMs” in the Neutron logical model, that would be huge. I’m hopeful we’ll finally nail the design and land code during Mitaka. In addition, the "work-around network segments" seems relevant to NFV deployments as well.

What are your key use cases for NFV?

Ford: Generally, our use case is to virtualize a network function (VNF) and automate the instantiation of the VNF, the network connectivity to it, and its life cycle (i.e. if an application moves, the network needs to ‘move,’ if it grows or reduces, the network needs to adjust). Specifically we’re looking at virtualizing basic connectivity, VPNs, security functions,
L4+, mobility services and connectivity, content management for video: all the core functions of AT&T.

Kim: Currently, our use cases for NFV are focusing on virtualization of the traditional telecom network functions such as IMS and EPC for elastic scale-out and control for service traffic explosion.

Our CTO, Dr. Alex Jinsung Choi, has a vision of an all-IT telecom infrastructure, which is trying to operate all of the telecom network functions on the cloud core in our software-defined data center. In the end, following the vision, all components in our core networks in data centers and local network operation centers will be running as virtualized network functions (VNFs) in cloud infrastructure with open source software such as OpenStack.

VNFs will be used for providing customer-specific dedicated telco services (multi-tenant) with orchestrated service chaining. These VNFs will also be used for enhancing telecom service quality and reliability with elastic VNF resource management and load balancing control.

The tenants might also include mobile virtual network operators (MVNO). The MVNO companies can provide better call quality with their own specialized, advantageous value-added services. This can also enrich the telecom cloud-based business ecosystem and make the adoption and use of OpenStack wide-spread.

Veiga: On the OpenStack team, NFV provides an opportunity to reduce the complexity of the core of our cloud architecture. With NFV, we can push functions and state closer to the edge of the network and free up the core to be as simple and fast as possible. We also gain the ability to virtualize access to networks, as well as virtualize and distribute managed services.

Superuser is always interested in thought pieces and other contributions from the community, please get in touch: [email protected]

Allison Price