Above, we argued that the 5G RAN and core will, with Releases 16 and 17, become capable of delivering the wide range of functions envisaged for ‘5G’. However, these will only turn into profitable new revenue streams and an agile commercial platform if operators also transform their operations, processes and the systems which surround the network itself.
One of the problems for early 5G deployers is that they had expected, a few years ago, that 5G would be inextricably linked with virtualized core and RAN. In reality, NFV (network functions virtualization) has come almost to a standstill amid criticisms that it is hard to scale, and as may operators turn away from first generation technologies and wait for fully cloud-native, containerized approaches to mature.
The risk is that the market will always be waiting for the next, better thing, creating a long hiatus as those emerging technologies harden sufficiently to be trusted in large-scale deployments. This has happened in the RAN – with commercial vRAN systems still at an early stage, and with many elements undeveloped, the first 5G roll-outs have sacrificed the economics and agility of that architecture, for a quick deployment.
Similarly, nearly all the 5G networks to be implemented in 2019 to 2022 will use the Non-Standalone radio and the 4G core, so operators will wait several years at least to introduce the virtualized 5G core. Some 4G cores are virtualized, and even cloud-native, but the number of operators which have transitioned to this architecture as their primary mobile core is small.
There is a small group of cheerleaders for the approach of getting the whole architecture right, and with clear commercial objectives in mind, rather than relying on a RAN and core upgrade to deliver everything. AT&T, Telefónica and NTT Docomo are obvious examples, and Turkcell has also been sliding into the fabled circle. For several years, it has been a pioneer in self-optimizing networks (SON) and virtualized evolved packet core (vEPC), both technologies that pave the way towards a grander transformation – but which have been adopted at scale by surprisingly few operators.
Now it claims to be building a “next generation network” which is designed, from the start, to be fully flexible in terms of the radio access technologies it will support (5G being just one option), and the use cases it will enable. It is working with Ericsson and Huawei on its RAN programs, but its overall focus is on developing platforms that are optimized for digital services, and therefore greater agility and new revenues. It has already launched some digital applications, such as non-SMS instant messaging, video streaming, and personal storage.
But these are conventional, and will find it hard to make much impact against the cloud majors. Once its platform is more comprehensive, it aims to be able to develop and activate new apps far more quickly in response to market demand – or even the needs of individual enterprises or user groups. In this way, it can innovate and experiment with lower risk or cost, and can differentiate with apps that are tailored to a particular requirement.
Of course, some operators have tried digital transformation and largely failed (Veon is an example), and others have found it far slower going than they had hoped (BT has acknowledged this, and even the mighty AT&T). Turkcell will face similar risks and disruption, but its overall approach does sound convincing.
It has placed RAN choices in second place behind flexible cloud infrastructure to support virtualization across the network, disaggregation of major monoliths like the RAN, and a high degree of automation. All this is being introduced based on its own set of software-defined operational models, and it plans to have 40% of its services virtualized by the end of 2019 (which will mark the end of the third year of the NFV project). It says it has virtualized 18% of its data and voice services capacity on its telco cloud so far.
The company has chosen the run-up to Mobile World Congress to trumpet its advanced state of transformation, and even offered some details. It has based its centralized virtualization infrastructure, which it calls Unified Telco Cloud, on Red Hat’s OpenStack Platform, with the same vendor’s implementation of Ceph for object storage, and with VNFs from Affirmed Networks and others.
Turkcell plans to unify all its services on this one cloud, which it believes will make it easier and quicker to implement new virtual network functions (VNFs) from different suppliers and to scale resources up and down to enable new applications. Affirmed, whose main product is a cloud-native mobile core, is providing the virtualized GiLAN (the interface between the RAN and the Internet), as well as a data optimization solution customized for Turkcell, and systems integration services. It will work with Turkcell for the next 3-5 years, managing day-to-day operations.
The resulting virtualized network already supports a wide range of functions, including:
- deep packet inspection (DPI) for traffic management
- data optimization for QoE differentiation
- carrier-grade NAT (CGN) for IP address management
- control plane services like PCRF and AAA for policy and charging
- IP Multimedia Subsystem (IMS) and Rich Communication Services (RCS) for voice and messaging including VoLTE.
These come from multiple VNF suppliers, showing that Turkcell is making progress towards every MNO’s dream – mixing and matching VNFs from different vendors in an interoperable, automated framework, with open orchestration.
“Our goal is to remake Turkcell as a digital services provider, and our business ambitions are global,” said CTO Gediz Sezgin said in a statement. “While planning for upcoming 5G and edge computing evolution in our network, we need to increase vendor independence and horizontal scalability to help maximize the efficiency and effectiveness of our capex investment.”
He also wants to leverage the cloud platform to create a rich ecosystem of partners and service providers. Sezgin continued: “With the Unified Telco Cloud, we want to lower the barrier to entry of our own network to make it a breeding ground for innovation and competition. In parallel, we want to unify infrastructure and decrease operational costs.”
The operator says the new platform reduces new service delivery time by at least three months, compared with its legacy network, and it has reduced cost by starting to eliminate proprietary hardware. According to Red Hat, Turkcell can now start to test new VNFs in 1-3 weeks, compared with three months or more with its legacy stack.