Yaniv Shaked, ASOCS
In the move toward virtualization in the cellular network, a novel approach uses a modem processing unit (MPU) and modem processing language (MPL) to alleviate the data transfer issues and speed up today’s overloaded cellular networks.
The Drive Toward C-RAN
The Cloud Radio Access Network (C-RAN) approach is gaining significant interest from cellular operators, cellular infrastructure vendors and integrated circuit (IC) vendors. With the C-RAN approach, the entire RAN processing is delegated to the cloud and implemented in a remote datacenter. The cellular site is composed of antennas and remote radio units (RRUs). The RRUs convert the wireless signals to and from digital baseband signals. The baseband signals are then transferred via a high-speed link (typically CPRI interface over optical fiber) to the datacenter. In the datacenter, a large number of baseband units (BBUs) handle the RAN processing. In the C-RAN paradigm, the entire processing chain—from baseband processing to packet processing—is performed by general-purpose servers.
The C-RAN approach provides significant advantages in capital and operational expenditures (CAPEX and OPEX) and facilitates advanced cooperative processing techniques. The CAPEX advantages are due to the fact that the C-RAN approach eliminates the need for expensive on-site equipment and real estate. The datacenter equipment itself is mostly comprised of general-purpose servers (such as Intelx86 servers), which is a mature and cost-optimized technology. The use of general-purpose CPUs allows the same processing platform to be used throughout the system, and thus the need for a separate platform for the physical layer and for other layers is eliminated. The generic nature of CPU processing platforms breaks the traditional bond between cellular infrastructure providers and cellular intellectual property providers. Thus, cellular operators are free to select the optimal cellular software for their needs. The centralized approach benefits from a statistical gain in which the datacenter resources need to be adjusted to the average of the expected traffic.
