Energy efficiency made easy

Energy saving and sustainability have become a top priority for service providers worldwide. Whether mandated to meet certain long-term net neutrality goals imposed by national climate action plans, or driven by soaring energy prices, service providers are seeking options to improve their energy footprint and establish a sustainable, environment-friendly operation.

Focusing on their network to reduce energy consumption is a logical choice, as a large portion of carbon emissions is generated from electricity consumed by mobile network base stations, communication station premises and data centers. Further increases are anticipated as the rollout of 5G continues and communications traffic volumes grow over time.

High energy-consuming network architectures

Conventional approaches for improving the energy footprint in data centers include adapting workload resources according to their demand by means of auto-scaling and dynamic scheduling. This approach is hard to implement in service provider networks where cloud-native network functions (CNFs) are typically deployed on specific CPU cores and given special access to system resources to maximize their performance. The 5G core UPF is one of these CNFs.

Polling and the use of the data plane development kit (DPDK) is a key mechanism used in the 5G UPF to deliver deterministic performance. By design, polling is oblivious to the load level of the 5G UPF. The associated CPU cores will always run at their highest utilization, even when the UPF is idle or receiving the lowest amount of traffic. The result is 5G UPFs always consume the maximum possible power.

Innovative reduction of energy consumption

To mitigate the energy impact of polling in DPDK workloads, Intracom Telecom has added the frequency feedback loop (FFL) feature to its NFV-RI™ solution. FFL dynamically selects the most efficient CPU frequency based on current demand, and is able to precisely determine how the traffic load of a DPDK-based CNF will vary in the very short term. FFL will adjust the frequency of CPU cores to reduce energy consumption without dropping packets. This creates opportunities for power savings and consequently reduced carbon emissions. 

In contrast to conventional energy-saving approaches, FFL can be applied transparently to a 5G UPF. No CNF modifications or the need to re-engineer deployment and traffic are necessary. There is no loss in performance as low latency is maintained and no packets are dropped, and FFL has minimal resource overhead (less than 50% of a single CPU core).

How Red Hat and Intracom Telecom can help

Red Hat and Intracom Telecom have collaborated to create a fully integrated and certified NFV-RI and Red Hat OpenShift solution to ensure streamlined deployment and stable implementation for service provider environments. Using NFV-RI on OpenShift enables 5G UPF workloads to take full advantage of power management capabilities to optimize overall energy consumption.

To accelerate the evaluation and adoption of the joint solution, Red Hat and Intracom Telecom have published a reference architecture on how to improve the energy footprint of the 5G UPF through an easily reproducible process. The reference architecture includes installation and deployment instructions, configuration options and Helm charts for the open-source components used as reference workloads (5G UPFs and traffic generators). 

The reference architecture also includes two common 5G data center use cases:

• Centralized 5G core deployments with fully-loaded UPF nodes
• 5G edge deployments with mixed-workload nodes

Both of these use cases are based on the open mobile evolved core (OMEC) 5G UPF distribution. In each, total savings in server power is reported for a real-world 24-hour traffic pattern. The fully-loaded use case achieves over 25% savings and the mixed-workload use case achieves over 15%.