Home > Communication Service Provider Solutions > Enabling Telecom Transformation > RAN Pooling - The case for RAN pooling with Cloud/Virtualized RAN > Improved energy consumption with C-RAN pooling
Traditional D-RAN deployment energy consumption increases with densification of RAN, especially for deployments using spectrum above 3 GHz. Hence energy consumption is a key focus area for RAN deployments. C-RAN pooling decreases energy consumption and costs, since fewer servers are needed when more NFs are shared. During periods of low traffic, BBU workloads can be moved to other servers and some servers can be switched off. Cooling and air conditioning costs are also reduced for pooled RAN, compared to traditional D-RAN. The more pooling and higher degree of shared NFs, the lower the energy consumption, making Centralized RAN pooling deployments the most energy efficient. However, the fronthaul, depending on the LLS option, may increase energy consumption. For example, split 8 with very few functions in RU incurs the highest power consumption on the fronthaul, while split 7.2 offers an optimal tradeoff – depending on the deployment requirements.
Algorithms that perform load-balancing of the incoming tasks across the servers running the NF workloads can impact energy efficiency gains. One such research [5] using modeling and simulation with 10 cells and the simulation parameters in Table 1, comprising maximum of 10 BBUs (for the distributed RAN case baseline) shows the gains as per Figure 8 and Figure 9.
Notations | Description | Range | Reference |
BW | Bandwidth | 10 (50 PRBs) | 20 (100 PRBs) |
Ant | Number of antennas | 1, 2, 3 | 1 |
K | Modulation | 1, 2, 4, 6 | 6 |
C | Coding rate | 1/3 – 1 | 1 |
The First Fit (FF), Next Fit (NF), and First Fit Decreasing (FFD) are bin-packing algorithms that use different methods to pack (load balance) the incoming traffic among servers. The number of computing servers can be reduced by matching the right amount of baseband processing with traffic load. For example, the FFD algorithm sorts the processing tasks from the RUs, packs the bigger traffic items first, and the smaller traffic items last. The NF algorithm does not order items based on remaining server capacity, so servers still have capacity after tasks requiring greater capacity are moved to the next server with that capacity. As the figures illustrate, the FFD algorithm achieves a nearly 73% savings in power consumption during low traffic periods.[5]
Furthermore, O-RAN Alliance is considering and defining specifications for Network Energy Savings (NES) [6]. Key use cases include workload evacuation from low-load physical nodes (servers), with the possibility of shutting down the low-load nodes to save energy (Figure 10). Vendor implementation of such O-RAN NES features will enable interoperability among multiple vendors supporting Cloud RAN pooling in the near future.