eBooks

radisys.com 10 Radisys Series — Who Disaggregated My RAN? eBook Meeting the Acceleration Challenge All this being said, the RAN still needs some specialized hardware and processing in the RAN base station functionality, especially certain functions in the PHY layer, fronthaul, and security and compression algorithms. PHY layer acceleration: Some of the PHY layer functional blocks including channel coding are compute intensive. These can be hardware accelerated using COTS components like FPGA, ASIC or GPUs. Such options are incorporated in base station and radio implementations. Fronthaul acceleration: In case of O-RAN fronthaul (split 7.2x) the benefits of compressions can be availed with hardware accelerating this compute intensive algorithm. Hardware acceleration can be achieved by FPGAs, ASICs or SmartNICs. Security and compression algorithms: Signaling messages are encrypted in the call flows and the processing can be offloaded to specialized hardware. And in the case of user data, IP header compression is another candidate for acceleration. Most often these are achieved by add-on pluggable cards on COTS servers. As the market for a decomposed base station using general purpose processors has grown, the vendors of ASICs, FPGAs and GPUs have recognized the growing opportunity to provide customizable acceleration functions. Because the accelerator market is already well developed, any company designing an open RAN solution can implement one of these hardware accelerators. Radisys' Open RAN 5G Software Suite supports hardware offload to the FPGA and can be uploaded to a hardware component where needed. Finally, open RAN vendors are able to leverage designs from the processors embedded in the servers. These chassis are readily available from multiple RAN vendors and allow operators to choose the performance, power requirements, and cooling requirements that they need. These white-box solutions enable operators to configure a wide number of base stations that meet their needs while reducing TCO. Cloudification of RAN Functions With all this decoupling of software and hardware in the RAN nodes, it is now easier for most of the RAN functions to be cloudified. As shown in Figure 1, these nodes on a cloud infrastructure are in the form of VMs or containers. There are some design challenges in cloudifying a DU in 5G against all the timing constraints with the various numerologies of 5G. Currently operators must deal with rather high operational costs due to proprietary network management systems and personnel required for RAN operations. In the traditional model, vendors provided operators with proprietary Element Management Systems (EMS) and Network Management Systems (NMS) management software of their own.

• Cover