Cloud-native technologies will be an essential part of next generation networks, able to deliver some of the full potential of 5G. Very few operators have deployed mainstream network functions in this way yet, but Australia’s Telstra has joined the elite group, implementing a cloud-native evolved packet core (EPC) that will support both 4G and Non-Standalone 5G services.
All operators so far have deployed 5G first in Non-Standalone (NSA) mode, which uses the 4G EPC and supports dual connectivity for devices to 4G and 5G base stations. Only relatively minor tweaks are required to the 4G core to migrate to NSA, but some operators are taking the opportunity to replace their legacy EPC with a new virtualized, more functional one, which can then be more smoothly migrated to a full 5G core when the move to Standalone (SA) 5G is ready.
In many cases, this upgraded EPC has been virtualized, with a path to cloud-native at the 5G SA stage, but Telstra is taking an early leap into the full cloud architecture, to improve its flexibility and functionality straight away. Rather than merely converting existing functions into software, as virtual machines, cloud-native involves designing them from scratch for the cloud environment, using containers and microservices, which can be aggregated and disaggregated, Lego-style, as required by changing application requirements.
Telstra is working with Ericsson to deploy its new vEPC on its NFV infrastructure (NFVi) and it says the system is already handling 4G and 5G NSA traffic.
Ericsson has supplied its Packet Core Gateway and its Packet Core Controller , the latter “deployed as a cloud-native container-based Mobility Management Entity (MME) in an existing MME pool”, said the vendor. Both products run on Ericsson’s Cloud Container Distribution (CCD), which is part of its NFVi solution, or on other distributions that conform to specs from the Cloud Native Computing Foundation (CNCF), which also runs Kubernetes, the dominant container management system.
Telstra said the core deployment is part of its T22 strategy to automate and simplify its network operations and service plans. “Through the T22 initiative, Telstra’s business is being transformed to improve service delivery and provide customers with enhanced experiences,” said Shailin Sehgal, the operator’s product enablement technology executive, in a press release. “To achieve this transformation, Telstra’s network needs to become more flexible and efficient, and cloud-native container-based applications are a key element of this. This is key to cost effectively scaling and automating our network and speeding up the delivery of innovative new services that are essential in a 5G world.”
Other ground-breaking operators in the virtualization area are not starting 2020 with such positive news. AT&T, considered by many to be the most comprehensive in its strategy to migrate its networks, processes and supply chains to the cloud environment, has nonetheless missed its own ambitious targets. Many telcos are cagey about discussing precise objectives for complex projects like network virtualization, with good reason, but AT&T has been very public about its aim of virtualizing 75% of its functions by 2020. This garnered it considerable profile and helped it to be seen as an industry leader in this area, which in turn helped it achieve its very influential position in cross-industry initiatives such as the ORAN Alliance. However, while Verizon’s almost equally advanced virtualization program may have flown beneath the awareness radar because of its less open approach to sharing milestones, at least it does not have to suffer negative comment when one of those goals is missed.
In fact, AT&T ended 2019 with 65% of its network functions virtualized, while Verizon was on 60%. This is still an impressive achievement, given the complexity of the task and the many new and unproven processes and technologies it involves. But it is well short of the 75% target AT&T set in 2017 (though it had already reduced that recently to 70%). The 75% deadline has now been shifted to the end of 2020, according to Scott Mair, president of AT&T technology and operations.
There is no shame in this. The companies which have been bold in taking the lead in virtualization also took a high degree of risk because they were dealing with so many unfamiliar factors. As AT&T’s CTO, Andre Fuetsch, put it in a conference address last year: “We left all the hard stuff for last”. Most operators engaged in virtualization have taken that approach, starting with relatively discrete and tested functions such as the 4G evolved packet core or the security gateway, and building up towards the harder challenges – and those which affect the whole end-to-end network – such as the RAN (which AT&T will not have virtualized even by 2021).
Rakuten, the trailblazer for greenfield end-to-end cloud-native deployment, has suffered its own setbacks and is a good example of how difficult it is to roll out such a platform in this very immature stage of the market.
However, Mair did announce another achievement, in a new year blog post. He said that “100% of the data traffic that runs through the infrastructure connecting the elements of our core network together is backed by SDN (software-defined networking)”.
The virtualization and SDN programs, which have been running since 2013, have many goals, one of them being to enable a scalable, flexible deployment of 5G. Key progress points have been to move the bulk of its data to run through MPLS tunnels under SDN control; to move its business applications to the public cloud (Azure), potentially paving the way to do the same for some network functions in future; to build its own network cloud, with a variety of partners such as Red Hat; and to shake up its supplier roster under its SDN 2.0 project.