How Carriers Build an NFVi to Support Network Transformation

In the NFV era, however, network services are migrated to the NFVi composed of COTS hardware and open-source software. Can the NFVi capabilities -- such as performance, reliability and security -- meet telecom service requirements? Will user experience deteriorate due to risks introduced by innovative technologies? The answers to these questions determine whether or not NFV can be commercially applied.

NFVi planners and designers are faced with the following question: What NFVi architecture is flexible enough to continuously meet service requirements? For example, can the NFVi constructed for 4G networks meet 5G service requirements? The answer to this question determines whether or not the NFVi can help carriers achieve their strategic goals for network transformation, and whether carriers can avoid the technical risks and wasted investment caused by planning issues.

However, NFV-related technical specifications have not been completely standardized so far. For example, the standard is not formally defined for the northbound interconnection between the NFVi and VNF/MANO. Currently, only the OpenStack API is introduced as a de-facto standard. This is likely to lead to risks in delivering projects with multi-vendor products and solutions, hindering the commercial use of NFV networks.

Performance
To achieve high performance, it is important to utilize a virtual network forwarding technology with high throughput and low latency. The performance of virtual switches not only determines whether the NFVi forwarding capabilities can meet commercial requirements for existing services (such as vEPC and vBRAS services), but also is crucial for user experience of new services (such as 5G and video services) over the NFVi platform. Improving performance also reduces carriers' costs and therefore has business value.

Since Moore's Law is running out of steam, it is not sustainable to rely on the performance improvement of general-purpose processors to continuously improve forwarding performance. SR-IoV, a hardware virtualization technology, provides an alternative to solve this problem. From a long-term perspective, software-virtualization and hardware-virtualization forwarding technologies will coexist for a long time in the NFV field, and play important roles in different service scenarios.

Reliability
Carriers should comprehensively analyze and harden the system reliability of multiple layers and fields to ensure that no single point of failure (SPOF) exists with a system and achieve a carrier-class service availability.

To be specific, the reliability of each component in the data, control, management and O&M planes in the NFVI should be hardened. Techniques, such as second-level alarms, automatic fault rectification on control components and VMs, automatic backup of service data and data center (DC)-level local and remote disaster recovery, should be used to fully ensure system reliability.

Security
As the industry transitions from closed, coupled telecom equipment to an open, decoupled NFV system, security is attracting more and more attention in the NFV field. In addition to the tenant-level isolation supported by computing, network and storage virtualization technologies, NFV1 security needs to be inspected and hardened from the following aspects: physical infrastructure security, software security, service data security and system security management.

Maintainability
Maintainability is also essential for commercial use of the NFVi. Since the NFVi is built using IT technologies, O&M tasks on the IT system, such as deployment, upgrades, patch installation, capacity expansion, alarms and fault management, are also involved in the NFVi, resulting in high technical requirements for the NFVi O&M system. For example, when Linux on physical hosts functioning as compute nodes requires an upgrade, automatically upgrading all the hosts in the same cluster without interrupting services is a big challenge. If such challenges cannot be overcome, the commercial use of the NFVI will be affected.

To overcome these challenges, an automated upgrade process management system needs to be added to the NFVi lifecycle management system, so that the sequence of service migration and node upgrades can be orchestrated in a programmed and automated way, which not only accelerates the upgrade process and lowers the upgrade risks, but also reduces the workload of the O&M team.

Meanwhile, diversified services require diversified resource types. Therefore, the NFVi must allow the use of heterogeneous computing, storage and network resources. With regards to computing, for example, VMs are required as basic computing resources and bare metal resources are required to support some performance-sensitive services. In addition to common x86 processors, high-performance heterogeneous computing resources (such as FPGA) are needed for computing-intensive services. Additionally, the NFVi must be able to integrate with an upper-layer container cluster management platform to support container-based services, such as agile 5G core network functions. However, the NFVi should remain decoupled from the platform and even the PaaS layer, preventing the NFVi architecture from becoming too complex and affecting NFVi transition and flexibility.