Distributed Ethernet Buffer Management for Congestion Control

The patent segments the common buffer pool into multiple distributed buffer areas across different Ethernet devices. Instead of using a single centralized buffer, the system divides buffering functionality across multiple nodes, each maintaining its own buffer space. This segmentation prevents any single buffer from becoming a bottleneck that reduces overall network throughput while maintaining high packet transmission speed through distributed parallel buffering.

The patent transitions from a single-dimension centralized buffer architecture to a multi-dimensional distributed buffer architecture. By adding the dimension of spatial distribution across multiple Ethernet devices, the system expands buffer capacity without creating a single point of saturation. This dimensional change allows packets to be routed through alternative buffer paths, maintaining throughput even when individual buffers experience congestion.

The patent segments the buffer system into multiple distributed buffers across different Ethernet devices. When one buffer experiences congestion, packets can be redirected to other distributed buffer areas, avoiding the time loss that would occur in a centralized system where all packets must wait for the single buffer to clear. This segmentation eliminates the single-point delay bottleneck.

The patent dynamically changes buffer allocation parameters based on real-time network conditions. When congestion is detected in certain buffer areas, the system adjusts packet routing parameters to redirect traffic through underutilized distributed buffer paths. This parameter adaptation prevents packets from being stuck in saturated buffers, thereby reducing transmission delays while maintaining high transmission speed through optimal path selection.

The patent segments the buffer capacity across multiple distributed Ethernet devices rather than relying on a single limited CBP. Each distributed buffer area contributes to the total buffer capacity, effectively expanding the system's packet holding capability without requiring any single device to have excessive buffer space. This distributed capacity prevents packet discarding by providing alternative storage locations when individual buffers are full.

The patent merges the buffering capabilities of multiple Ethernet devices into a unified distributed buffer system. By combining the buffer resources of multiple nodes, the system achieves a total buffer capacity that exceeds what any single device could provide. This merged distributed capacity ensures that packets are rarely discarded due to buffer saturation, thereby maintaining packet delivery reliability and preventing out-of-order reception caused by packet loss.

The patent implements a universal distributed buffer management protocol that can be applied across different Ethernet devices regardless of their specific hardware characteristics. This universal approach allows multiple devices with varying buffer capacities to participate in the distributed system without requiring device-specific customization. The standardized protocol simplifies management complexity by providing a uniform interface for packet routing and buffer allocation across the entire distributed network.

The patent incorporates feedback mechanisms where Ethernet devices continuously exchange buffer status information with each other. Based on this feedback, the system dynamically adjusts packet routing decisions to maintain optimal throughput. The feedback loop provides real-time information about buffer availability, allowing the network to automatically redirect traffic away from saturated buffers without complex manual configuration. This self-regulating feedback system manages complexity by using simple local decisions based on global status information.