With 5G growing exponentially, service providers are investing billions of dollars in their network infrastructure to meet demand for a broad range of emerging services. A key challenge in addressing the explosive growth of 5G is deciding where to allocate computing capacity in their networks to serve the unique performance requirements of these diverse services.
Service provider’s core and edge data centers need to meet 5G computing capacity demands for ultra-low latency and high throughput services. New innovations in the design of the 5G user plane function (UPF) can help them enhance their packet processing capabilities and satisfy demand through efficient management of high data usage applications.
The UPF is the workhorse for the flow of digital data traffic, with design and changes in standardization allowing for the disaggregation of the network’s control and user plane for enormous efficiencies. However, the UPF could become a bottleneck if too many packets flow through the network at a time. To eliminate this requires a different approach using innovative software technologies.
Ensuring consistent performance
Additional burden on the user plane that raises throughput to multiple terabytes per second has primarily been addressed with higher capacity hardware. This approach can meet increasing demand, but is unable to efficiently manage the complexity of associated software and protocols. With a growing range of applications at the edge that mandate the placement of UPFs closer to customer locations, service providers will fail to deliver a consistent performance if they rely upon data processing hardware alone.
To address this challenge Red Hat, Intel and Kaloom have collaborated to provide an edge-based UPF that can deliver 2.2 Tbps (full duplex) throughput for over 4 million subscribers using a heterogeneous hardware acceleration strategy. The technical collaboration between the three organizations exemplifies the latest innovations in UPF packet processing with Kaloom’s Unified Edge Fabric™, Red Hat’s OpenShift™ on P4-programmable switches, based on Intel® Tofino™ Intelligent Fabric processors , Intel’s Stratix-10 FPGAs and Intel Xeon processors.
This innovative design improves performance at a lower cost while significantly reducing network latency from milliseconds down to microseconds. This allows service providers to meet the demands of 5G-enabled applications such as the internet of things (IoT), augmented and virtual reality (AR/VR), and many more.
An innovative UPF software design that uses heterogeneous programmable hardware allows service providers to innovate faster and introduce new services to meet evolving requirements, in terms of higher bandwidth, lower latency, and subscriber scalability. Red Hat OpenShift provides a cloud-native application platform and deploys the entire solution into a compact footprint. Intel® Tofino™ Intelligent Fabric processors, Intel’s Stratix-10 FPGAs and Intel Xeon processors ensure higher throughput, lower latency, and an improved cost-performance ratio compared to traditional server deployments.