NTT DoCoMo is poised to scale up its NFV implementation after reporting increased efficiency and lower costs. It comes after the operator rolled out the world’s first multi-vendor ETSI NFV commercial system in March 2016.
Running on Ericsson-provided, OpenStack-based virtualized infrastructure managers (VIMs), DoCoMo’s implementation includes virtual network functions (VNFs) from multiple vendors such as NEC, Nokia, and Fujitsu.
The VNFs are running on an NFV infrastructure (NFVI) consisting of commercial off-the-shelf (COTS) hardware and SDN switches. The operator automates VNF lifecycle management in a standard way, using an NFV orchestrator, VNF managers, virtualized infrastructure managers (VIMs), and NFVIs complying with ETSI NFV architecture and specifications.
The benefits are already becoming clear. Because it takes advantage of COTS hardware, the implementation has reduced both capex and maintenance expenditure. At the same time, an auto-scale out feature to accommodate more traffic on-demand allows the operator to utilize improved connectivity to support a larger number of user sessions during congestion.
DoCoMo’s NFV deployment also facilitates a shorter time to market, with service releases now available in a day, rather than in a few months. Meanwhile, NFV has enabled increased automation in service delivery, also saving time.
This is in addition to quality improvements. Using management and orchestration (MANO), the operator is able to reduce real-time, on-site failure recovery and maintenance events. DoCoMo is hoping the ETSI NFV-compliant OPNFV (Open Platform for NFV) Doctor will improve this feature even further.
In order to achieve the benefits from its deployment, DoCoMo made minor modifications in legacy operations support systems (OSS) and network functions, based on ETSI NFV MANO specifications.
DoCoMo now wants to reduce opex further and avoid human error by automating VNF and network configuration.
A Successful Deployment, and the Challenges
The deployment has so far been a success, but DoCoMo faced several challenges when implementing NFV. Indeed, the area is particularly complex: evolved packet core (EPC) VNFs consist of many components, requiring numerous internal networks to be set up.
Therefore, DoCoMo had to use manual resource reservation and allocation based on elements such as desktop planning and single root input/output virtualization (SR-IOV) to meet its virtual EPC requirements.
At the same time, DoCoMo lacked the finer root-cause analysis tools needed for decoupling software from hardware. The operator therefore had to maintain detailed mapping between virtualized and physical resources, noting the location of virtual machines and how they were connected.
Meanwhile, one of the key challenges in NFV is management and orchestration. Because VNFs consume infrastructure resources in different ways, DoCoMo needed to create automated, per-VNF operation procedures to reduce errors in the processes.
However, DoCoMo points out that differences in VNF operation procedures are expected to disappear once vendors develop their products based on ETSI NFV standards.
Even so, these are useful lessons for every operator implementing NFV. It is with this in mind that DoCoMo is continuing to bring its more than two years of NFV implementation and operation experience as contributions to ETSI NFV standards. DoCoMo will continue to do so based on its NFV commercial system to make ETSI NFV standards and its consequent commercial implementations more robust, practical, and stable.
Skills for NFV
The industry is learning from NFV deployments, but more work still needs to be done. In order for virtualization to reach its full potential, DoCoMo expects to see some deployment related issues addressed and solved.
In the future, DoCoMo sees a need for the ETSI NFV specifications to take into account the formidable task of having multi-vendor VNFs running on an OpenStack-controlled cloud and NFVI, yet ensuring that maintenance can be performed continuously, since this is a common situation for large nationwide networks.
In the meantime, it is essential that companies have the skills to take on the brave new world of NFV. Virtualization offers benefits, but many of these can only be achieved once sufficient expertise is available in-house.
DoCoMo has learned that in order to reap the full benefits of virtualization, telecom operators should deploy NFV sooner rather than later. This will allow them to acquire the know-how on the cloud and related technologies, as well as of new ways of working, such as DevOps.
At the same time, it is important that the industry considers the ecosystem necessary for the wider adoption of NFV, which will include many open source solutions. Indeed, open test beds will play a major role in the development and testing of different NFV components including MANO. And these will be indispensable in achieving high-quality VNF and MANO products in a faster and more cost-efficient way.
It is also important to take into account the different life cycles of components, and the fact that they will need upgrading at varying times. ETSI NFV-defined reference points and interfaces define the interoperability of the NFV functional blocks, which helps operators update each component more independently. Further work is expected, especially in the product and solution development area, to ensure fully independent component upgrades without service disruption or complex and time-consuming processes.
Boosted by ETSI’s continuing work on standards, DoCoMo’s deployment is expanding. 75 percent of its mobile core network assets will be running an ETSI NFV based virtualized system by 2020.
After that, the operator will introduce enhancements to the current deployment based on the standardized ETSI NFV Release 3 features, such as inter-datacenter connectivity services supporting multi-site network services. Core network applications will become cloud-native as this technology comes together to become the supporting baseline for 5G services.
As NFV deployments accelerate, DoCoMo predicts an increasing number of compliant VNFs and operation support systems will soon be available in the market. This will be further enhanced by operation automation including artificial intelligence, which will evolve NFV into an even smarter autonomous system.
This article was co-authored by Yuya Kuno, engineer of the network management and orchestration group of the core network development department in NTT DoCoMo.