Virtual Output Queue Latency Control With Adaptive Tail Drop
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network devices face limitations in packet forwarding due to finite resources, leading to increased latency and undesirable consequences such as packet resending or discarding, which can be mitigated by managing virtual output queue latency and proactively dropping packets based on latency states.
Innovation Solution
A system that monitors latency states of virtual output queues and allocates resources to reduce buffering time, proactively dropping packets when necessary to meet predetermined latency goals, thereby avoiding deleterious downstream impacts and optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If packets are buffered in virtual output queues to maintain network flow, then packet forwarding capability is improved, but latency increases and resources are consumed
Solution Approach 1:
The patent implements dynamic latency management by continuously monitoring the latency state of virtual output queues and adjusting buffering behavior in real-time. The system transitions between different buffering modes based on current latency conditions, making the packet forwarding system adaptive rather than static. This resolves the contradiction by allowing high buffering capability when latency is acceptable while reducing buffering when latency becomes excessive.
Solution Approach 2:
The system changes the latency parameter of virtual output queues dynamically based on monitored conditions. By adjusting latency thresholds and buffering parameters in response to real-time queue states, the system optimizes the balance between maintaining packet flow (productivity) and controlling buffering time (latency loss).
2Loss of time
If buffering resources are increased to reduce latency, then latency is improved, but resource consumption increases
Solution Approach 1:
The patent employs dynamic resource allocation where buffering resources are allocated to virtual output queues based on real-time latency conditions and traffic demands. Rather than statically increasing resources for all queues, the system dynamically adjusts resource distribution, allocating more buffer space to queues experiencing high latency while maintaining efficient resource utilization across the network device.
Solution Approach 2:
The system applies different buffering resource allocations to different virtual output queues based on their specific latency conditions and traffic characteristics. Each queue receives localized resource optimization rather than uniform resource distribution, allowing efficient use of total buffering resources while addressing latency issues where they specifically occur.
3Productivity
If packets are proactively dropped based on latency states, then downstream processing efficiency is improved, but packet loss increases
Solution Approach 1:
The patent implements proactive packet dropping as a preliminary action to prevent downstream processing of packets that would inevitably cause problems. By monitoring latency states and dropping packets before they reach problematic buffering conditions, the system prevents downstream congestion and processing inefficiencies. This resolves the contradiction by sacrificing some packets (controlled loss) to protect overall downstream processing efficiency.
Solution Approach 2:
The system takes preliminary anti-action by dropping packets that are likely to cause downstream issues before they can negatively impact the network. This preventive approach counteracts potential downstream congestion and processing failures by removing problematic packets early in the forwarding path, thereby maintaining overall network productivity.
Data Source
AI summary
A network device includes a switching system for directing packets between ingress ports and egress ports of the network device. The network device also includes a switching system manager that makes an identification of a state change of a virtual output queue of the switching system; and performs an action set, based on the state change, to modify a latency of the virtual output queue to meet a predetermined latency in response to the identification.