Resource pooling between BBU frames means that resources can be allocated among BBUs. Compared with pooling between baseband boards, pooling between BBU frames is suitable for larger-scale pooling scenarios.
Similar to the inter-board data interaction scheme, from the perspective of interaction complexity and energy saving, BBUs are suitable for exchanging cell-level data. There are three BBU architecture evolution schemes to realize data interaction between BBU frames.
Evolution architecture 1: Forward data through the baseband board. Inside the baseband board, all fronthaul optical modules can be connected through high-speed plug-in buttons or switching chips. After the fronthaul data is transmitted through the optical module, it is forwarded to other optical modules, and then transmitted to other BBUs via optical fibers.
Evolution architecture 2: Forwarding through the switching board in the BBU. This architecture is consistent with the evolution architecture designed for inter-board data interaction. After the fronthaul data is transmitted from the optical module, it is forwarded by the switch chip in the switchboard, then output through other optical modules, and transmitted to other BBUs through 100G QSFP28.
Evolution architecture 3: Design independent switching equipment. This method breaks the structure of direct connection between BBU and AAU but is connected through switching equipment. The switching equipment has a built-in routing-controllable switching chip, and the two sides are optical modules. The module, after the photoelectric conversion, is forwarded to the corresponding link by the switching chip, and then the photoelectric conversion is realized through the optical module on the other side of the switching device, and transmitted to the BBU through the optical fiber. This architecture is more conducive to large-scale data interaction between BBUs, and can adjust the data flow direction of antennas in real-time according to configuration commands, improving the scalability and flexibility of BBU base stations.
The comparison and analysis of the three BBU architectures from the four dimensions of performance, energy consumption, cost, and reliability are as follows.
Performance: When architectures 1 and 2 are used to implement pooling between BBU frames, unlike baseband boards, the pooled data can be transmitted through the backplane. At this time, the data needs to be forwarded through optical fibers, and the optical fiber connection is usually fixed. Therefore, architecture 1 2. When implementing pooling between BBU frames, the interaction method lacks flexibility. Architecture 3 is very flexible by controlling the routing and forwarding of switching devices. Architecture 2 occupies an independent BBU slot, and Architecture 1 and 3 do not involve this issue. In architecture 3, two optical modules and one switching chip are added to the transmission link between the AAU and the BBU, a total of three nodes, and the system delay increases, so the software scheduling scheme needs to be optimized.
Energy consumption: After adopting architectures 2 and 3 to implement inter-frame pooling, the original baseband board or BBU can be in deep sleep or shutdown state, while architecture 1 still needs to retain the switching function and monitoring function after implementing inter-frame pooling.
Cost: The switching chips of Architecture 2 and 3 need to be designed for full load conditions, and a higher switching chip cost is required. In addition, the demand for optical modules in architecture 3 increases exponentially. Therefore, the cost relationship of the three architectures is Architecture 3 > Architecture 2 > Architecture 1.
Reliability: Both architectures 2 and 3 involve centralized switching issues. The centralized switching scale of architecture 3 is larger, facing the possibility that multiple devices may be paralyzed when a switching device fails. But relatively speaking, when adopting architecture 3, all BBUs connected to the same switching device are mutual backups. When a certain BBU or a certain baseband board fails, other baseband boards can complete the cell processing by controlling the switching route.
Through the comparison and analysis of the three evolutionary architectures through the above four dimensions, it can be found that when the interaction between BBU frames is large, architecture 3 has the advantage of strong flexibility, while architectures 1 and 2 are flexible due to the fixed optical connection relationship. Sexually restricted. However, architectures 2 and 3 have high costs and relatively low reliability, and architecture 3 still needs to solve a larger system delay. Although the energy-saving ratio of architecture 1 is lower than that of architectures 2 and 3, it does not need to sacrifice the space of the baseband board and the cost is lower. At the same time, the risk of failure is more dispersed. When the scale of exchange between BBU frames is small, it is a more preferred evolution architecture.