Dynamic Input Port Scheduling for Buffer Overflow Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for scheduling data communication input ports fail to prevent buffer overflow and packet loss due to inconsistent bandwidths and insufficient polling rates, leading to suboptimal performance in network processors.
Innovation Solution
A device and method that dynamically allocate buffer sizes and service time slices based on priority and bandwidth, using a resource preprocessing module to set priorities, a port state monitoring module to classify active ports, a buffer resource allocating module to allocate buffer sizes, and a polling port module to poll ports from high to low priority, ensuring efficient resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If equal time slice polling is used for scheduling input ports, then each port receives equal service opportunity, but ports with different bandwidths experience buffer overflow when their traffic exceeds the fixed polling rate
Solution Approach 1:
The patent implements dynamic polling periods for each input port based on its bandwidth characteristics. The polling period is calculated as Total Polling Period × (Port Bandwidth / Total Bandwidth), allowing ports with higher bandwidth to be polled more frequently. This dynamic adjustment prevents buffer overflow by matching the polling rate to the actual traffic generation rate of each port, while adapting to different bandwidth configurations.
Solution Approach 2:
The patent applies different polling strategies to different input ports based on their local characteristics (bandwidth). Each port receives a customized polling period proportional to its bandwidth, rather than a uniform polling period. This local quality approach ensures that each port's buffer size and polling rate are optimized for its specific traffic requirements, preventing overflow at high-bandwidth ports while maintaining fairness.
2Productivity
If fixed buffer sizes are allocated to input ports, then resource allocation is simple and predictable, but ports with higher bandwidth requirements experience packet loss when traffic exceeds the fixed buffer capacity
Solution Approach 1:
The patent dynamically adjusts the buffer size parameter for each input port based on its bandwidth. The buffer size is calculated as Total Buffer Size × (Port Bandwidth / Total Bandwidth), ensuring that ports with higher bandwidth allocations receive proportionally larger buffers. This parameter change allows the system to handle varying traffic loads at different ports without packet loss, while maintaining efficient use of the total buffer resource.
3Reliability
If polling rate is increased for all ports to prevent overflow, then buffer overflow is prevented, but the processing capability of the single processor becomes insufficient and overall system performance decreases
Solution Approach 1:
The patent applies differentiated polling rates to different input ports based on their individual bandwidth requirements. Instead of uniformly increasing the polling rate for all ports, each port receives a polling period proportional to its bandwidth (Total Polling Period × Port Bandwidth / Total Bandwidth). This local quality approach ensures that only ports with higher traffic demands are polled more frequently, preventing buffer overflow at those specific ports while maintaining processor efficiency overall.
Data Source
Figure 1~2
Figure 3
AI summary
The present invention provides a method and device for scheduling data communication input ports, which comprises: monitoring a state of input ports, acquiring the information about valid input ports, dynamically allocating the buffer size for the valid input ports according to the valid input ports, calculating a time slice according to the bandwidth and priority of each valid input ports, and serving each valid input port according to the priority of the valid input ports and different time slices. With the solution of the present invention, the problems of buffer overflow and packet loss caused by the inconsistent bandwidths of the valid input ports and insufficient polling rate are eliminated, thereby improving the performance of processors.