Aggregate Policing via Probabilistic Filtering for Fair Bandwidth Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network switching devices face challenges in managing data packet flow from multiple sources to prevent resource depletion and congestion, as existing methods either fix input bandwidth or implement proportional fairness, leading to unfair bandwidth distribution and potential starvation among sources.
Innovation Solution
The implementation of probabilistic filtering within network switching devices calculates data packet arrival rates and guaranteed shared bandwidth rates for each source, selectively passing or dropping packets based on calculated probabilities to ensure fair distribution according to the destination's bandwidth capacity, avoiding fixed bandwidth allocation and proportional fairness issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If aggregate policing is implemented to limit total bandwidth to destination, then destination congestion is prevented, but individual source fairness is compromised
Solution Approach 1:
The patent segments the aggregate policing function into per-source rate calculation and probabilistic filtering components. Each source's traffic is independently evaluated against its calculated rate, enabling fair individual treatment while collectively preventing destination congestion through aggregated control.
Solution Approach 2:
The system implements feedback by continuously monitoring traffic rates from multiple sources and dynamically adjusting probabilistic filtering decisions. The calculated rate for each source serves as feedback to determine packet acceptance probability, ensuring both individual fairness and collective congestion prevention.
2Reliability
If fixed input bandwidth is assigned to each source, then resource depletion is prevented, but bandwidth utilization efficiency decreases
Solution Approach 1:
The patent transitions from static fixed bandwidth allocation to dynamic rate calculation. Each source's allowable rate is continuously determined based on current traffic conditions and destination capacity, enabling both resource protection and efficient utilization by adapting to changing demands in real-time.
Solution Approach 2:
The system changes the parameter of bandwidth allocation from fixed to dynamically calculated values. By computing individual source rates based on aggregate capacity and traffic patterns, the system optimizes both resource protection and utilization efficiency through parameter adaptation.
3Productivity
If proportional fairness is implemented among sources, then bandwidth distribution is optimized, but starvation of low-rate sources occurs
Solution Approach 1:
The patent introduces probabilistic filtering as an intermediary mechanism between proportional fairness allocation and actual packet transmission. This intermediary protects low-rate sources by providing probabilistic protection, ensuring they receive minimum service while still allowing high-rate sources to utilize available capacity efficiently.
Solution Approach 2:
The system creates a protective layer by copying the traffic flow through a probabilistic filtering stage. This copying mechanism ensures that low-rate sources are protected from starvation while maintaining the overall proportional fairness distribution, as the filter independently evaluates each packet against source-specific rates.
4Ease of operation
If probabilistic filtering is applied per source, then source fairness is improved, but device complexity increases
Solution Approach 1:
The patent implements a universal probabilistic filtering mechanism that handles multiple sources simultaneously through a single destination-focused control structure. This multi-functional approach maintains source fairness while avoiding the complexity of separate filtering systems for each source, as the same filtering logic applies to all sources based on their calculated rates.
Data Source
AI summary
In one embodiment, a method comprises calculating a corresponding data packet arrival rate for each of a plurality of data sources supplying data packets destined for a prescribed destination, the prescribed destination within a machine and the prescribed destination having a bandwidth capacity; calculating a guaranteed shared bandwidth rate for each data source based on assigning a corresponding selected portion of the bandwidth capacity relative to the corresponding data packet arrival rate; selectively passing each data packet from the corresponding data source as a passed data packet, or dropping the corresponding data packet, according to a calculated probability that the corresponding data packet arrival rate does not exceed the corresponding guaranteed shared bandwidth rate; and selectively filtering the supply of aggregated passed data packets, aggregated only from among the passed data packets supplied by the data sources, to the prescribed destination according to the bandwidth capacity of the prescribed destination.