There may not be a business case for rolling out 5G yet, but it’s better to be an early adopter if only to get ahead of 5G’s steep learning curve, according to Telstra technology team manager Jeff Schmidt.
Speaking at Apricot 2017 event in Ho Chi Mihn City, Schmidt described how Telstra’s current network has physical nodes and a lot of manual configurations with minimal opportunities to differentiate traffic type with QoS. Going forward with its new 5G network Telstra is introducing the concept of network slicing where it runs nodes like its PAC gateway in a virtualized environment, with dedicated computing and storage resources to those elements with the aim of allowing differentiate services based on each network slice.
All this has to be done in a 5G network that has a target of 1ms latency across the network which makes it all the more challenging, although with NFV, there is the potential to regionalize the network across Australia and reduce latency by offering more services regionally.
Such differentiation could be things like mobile broadband, voice, OTT video, mission critical communications and massive IoT to name but a few, Schmidt said.
Mobile edge computing (MEC) is used to run services close to the base station in an NFV environment, pulling some services out of a cell site and moving them into a local site servicing the network rather than to a centralized datacenter. All of this means that the bottom line is a better customer experience.
Schmidt said Telstra is also working on horizontal SDN slicing of the network, which has more to do with radio interface and improving capacity, by putting more compute services onto the network. This means that as 5G evolves, it enables the ability to select a different SDN slice of the radio network as well as the carrier network. A corporate VPN could, for instance, have different rules on the radio network as well as the data network.
Schmidt noted that the 4G and 5G cores are separate, and while slicing is an evolution of 4G, the decision to go with a separate core with NFV for 5G was to help with the building of skills within the organization, and to enable Telstra to be one of the earliest adopters. With the new 5G network, what is becoming clear is that the data center is extending into the RAN and a different set of skills are needed for the engineers. A whole new set of challenges are needed in orchestrating the spin-up of a new packet gateway, for instance.
Each network slice may have sub-slices that might have separate packet gateways and separate policy control platforms – but at some point they all have to interact with each other.
Along the way, there are many problems, he acknowledged. For instance, in the new architecture, both the packet gateway and the carrier-grade NATs are virtualized. One vendor is offering 8 Gbps of throughput per VM, which can be extended to 40 Gbps but at the cost of losing the ability to move traffic around. Scale is a big challenge, as is the need to implement IPv6 to cope with demands that are expected to be at least an order of magnitude greater than the current 4G network loads, he said.