Virtualizing service provider workloads calls for a thorough and thoughtful plan that taps into the full ecosystem, from infrastructure orchestration to the application itself. At OpenStack Summit this past May, Julio Villarreal-Pelegrino, principal architect with Red Hat’s cloud practice, and Stephane Lefrere, head of the cloud infrastructure practice at Red Hat, shared an approach for creating a successful NFV architecture, with a focus on the platform features and tools that support service assurance, performance, manageability and operability.

The presenters kicked off the session by asking attendees to describe in one word what network functions virtualization (NFV), means for them. A real-time word cloud showed a top result of magic, as well as words like automation, time-to-market, provisioning and virtualization.

What does this “magical” concept of NFV really mean? The presenters defined NFV as “decoupling of network functions from underlying physical network infrastructure” and “move of traditional network functions usually deployed in proprietary hardware to software running in virtual machines (VM) on general-purpose hardware or cloud infrastructure.”

Based on an audience poll, lack of flexibility was determined the biggest driver for NFV adoption for attendees at the session, with high cost coming in second. Vendor lock-in and slow innovation were also noted by attendees as being important. These are all big issues for telecommunications businesses, and with a quick show of hands Villarreal-Pelegrino confirmed that most attendees in the room worked in the telco space.

Lefrere pointed to legacy infrastructure as a key reason that NFV adoption is on the rise. There are a lot of legacy systems out there representing hundreds of millions of dollars in investment, he says, and the lack of flexibility and the obligation of IT consumers to go through the network team to set rules is the old way of doing things. “Lack of flexibility is an impediment for businesses to respond to the demand,” continued Lefrere.

In order to be able to scale and respond to high traffic loads, you need to be able to be elastic around workloads, both on the compute and network side. Providing on-demand and self-services allows developers to innovate faster, and open standards also open the door for driving faster innovation.

Lefrere then walked through an overview of benefits NFV delivers:

  • Lowers costs: reduces CapEx to eliminate wasteful over-provisioning and reduces OpEx because many of the space, power and cooling costs go away with virtualization
  • Increases flexibility: management automation and reusable infrastructure
  • Empowers innovation
  • Easy to scale
  • Faster time-to-market

The presenters pointed to data from a 2014 Doyle Research report that 83 percent of telco operators demand or prefer open systems for their networks and 95 percent of telco operators see open source as a positive attribute for NFV solutions.

But why open standards? According to the presenters, open standards allow companies to avoid vendor lock-in and to take charge of their own destiny, some of paramount importance to telco operators.

Lefrere provided an overview of what NFV architecture looks like, saying, “the idea with this OpenStack platform is to be able to manage your compute, resource, networks, storage and have that as the base platform for innovation on the VNF side.”

The presenters paused to get a snapshot of OpenStack use from the audience. Their real-time survey showed 60 percent of the audience was using OpenStack for NFV workloads and 40 percent was not. Users used the words cost, rapid, open and flexibility as key reasons for using OpenStack.

OpenStack offers a number of advantages over traditional virtualization, said Lefrere, including modular components, multi-tenancy, pluggable storage and networking, a rich API feature set and a vibrant community.

The presenters also reviewed the capabilities delivered by individual modular OpenStack components to meet Virtualized Infrastructure Manager (VIM) requirements. Villarreal-Pelegrino started by reviewing OpenStack’s control and management of compute, networking and storage resources through allocation and release of individual resources, mapping physical resources to virtual, secure resource sharing across multiple tenants and quota management and enforcement. Villarreal-Pelegrino went on to say OpenStack also provides platform resiliency with a high available control plane, exposes APIs to enable orchestration across services and domains, and provides mechanisms to collect fault and performance data for physical and virtual resources.

Another crucial part of NFV deployment is the software-defined network (SDN). In fact, according to Villarreal-Pelegrino, “the SDN conversation is probably the most important conversation you’re going to have when you are onboarding OpenStack as an NFV in your company, because everything that’s in there will be dependent on that SDN choice.”

Red Hat SDN partner Neutron was highlighted as an option that is fully integrated into OpenStack development and maintenance tools and cycles. Neutron provides out-of-the-box support for VLAN, VxLAN and GRE overlays as well as open source that’s supported by the upstream and utilizes upstream components. Neutron’s SDN solution includes quality of service (QoS) and performance optimization.

Storage is also key, and the presenters discussed both open source and commercial alternatives. One popular option is CEPH storage, which is open source, highly scalable, reliable and has built-in redundancy. CEPH utilizes commercial, off-the-shelf x86 hardware and standard Ethernet and IP connectivity, and is supported as a backend for all OpenStack storage components such as Cinder, Glance, Swift, and Nova ephemeral.

The presenters cautioned that if you use an alternative commercial storage solution, you need to make sure the solution is certified. They offered a list of elements to consider, and questions to ask, if choosing a commercial SDN or storage provider:

  • Is the provider vendor-dependent?
  • Does the provider utilize proprietary components?
  • Will the solution require integration with NFVi deployment tools?
  • Does the storage solution have different development cycles from other NFVi components?

According to Villarreal-Pelegrino, SDN providers may also provide specialized features and have potential faster feature implementation and customization support, and on the storage front, specialized connectivity requirements.

To wrap up, the presenters talked about aiming for an optimized platform to achieve end-to-end service performance through performance optimization of individual components and platform-aware service placement. They pointed to using existing features that already exist in the Linux kernel to make them production-ready for NFV.

Villarreal-Pelegrino covered monitoring and troubleshooting tools, reviewing Open Source options like Elasticsearch, Fluentd, Kibana and Nagios as well as integration with existing tools like SNMP and syslog. He also reviewed use cases and stated the importance of linking your use case to the solution. “Do not adopt OpenStack because it’s the cool, trendy thing,” he says. “Adopt OpenStack because you actually have a use case that you can fix and match it with OpenStack.”

Considering the three areas of compute, networking and storage, the presenters reviewed Red Hat OpenStack Platform’s NFV features and highlighted Red Hat Ceph Storage, which is bundled with Red Hat OpenStack Platform and offers complete storage for OpenStack. They also provided an overview of Red Hat CloudForms, “which is crucial in our approach to manage OpenStack,” said Villarreal-Pelegrino. The presenters touched on automation, with a focus on Ansible’s simple, powerful and agentless solution, and then put all of the pieces together for a review of a full stack use case that’s ready for deployment.