Network Device Packet Sampling Rate Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Time-based sampling in network devices often results in oversampling due to not accounting for the actual packet rate, leading to reduced accuracy of network traffic characteristics and unnecessary resource burden.
Innovation Solution
Implementing a method that compensates for time-based sampling by packet elimination, where packets are replicated, and based on a configured sample rate and current sample rate, packets are either queued or dropped to maintain the desired sampling rate, using a credit refresh scheme and identification hash keys to manage sampling across multiple monitoring points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If time-based sampling is used to simplify implementation, then device complexity is reduced, but measurement precision deteriorates due to oversampling
Solution Approach 1:
The patent implements feedback by continuously monitoring the actual packet rate and comparing it with the configured sample rate. Based on this comparison, the system dynamically adjusts the sampling behavior through sample credits, ensuring that the desired sample rate is maintained while simplifying the implementation compared to complex adaptive sampling algorithms.
Solution Approach 2:
The system changes the sampling parameter dynamically by using sample credits that are allocated based on the configured sample rate and consumed based on actual packet sampling. This parameter adjustment mechanism allows the system to adapt to varying packet rates while maintaining a simple time-based sampling implementation structure.
2Ease of operation
If time-based sampling is used to achieve simple implementation, then ease of operation is improved, but loss of substance increases due to unnecessary packets being sampled
Solution Approach 1:
The feedback mechanism monitors the relationship between configured sample rate and actual sampling rate, adjusting packet acceptance through sample credits to prevent unnecessary sampling of packets when the actual rate is below maximum expected rates, thus conserving network resources while maintaining operational simplicity.
Solution Approach 2:
The sampling system serves itself by using the information from packet rate monitoring to automatically adjust its own sampling behavior through the sample credit mechanism, eliminating the need for external control while optimizing resource usage and maintaining simple implementation.
3Productivity
If time-based sampling samples N packets in every predetermined time interval, then productivity is improved by maintaining consistent sampling rate, but measurement precision deteriorates when actual packet rate is less than maximum
Solution Approach 1:
The patent introduces dynamics by making the sampling process adaptive through sample credits. While the time-based sampling structure remains static and simple, the actual sampling behavior becomes dynamic as packets are selectively accepted or rejected based on the comparison between configured and actual sample rates, improving precision without sacrificing productivity.
Solution Approach 2:
The system changes the effective sampling parameter by using sample credits to adjust the number of packets sampled from the predetermined time interval. This allows the system to maintain consistent time-based sampling operation while dynamically adjusting the actual sampling rate to match network conditions, resolving the precision-productivity contradiction.
Data Source
AI summary
Methods and apparatus for compensating for time-based sampling using packet elimination are disclosed herein. An example method may be implemented in a network device communicatively connected to a network. The method may include: receiving a packet over the network at a monitoring point; replicating the received packet; determining whether to sample the replicated packet using a time-based sampling scheme; upon determining that the replicated packet should be sampled, queuing the replicated packet to a port of the network device; determining whether to eliminate the queued packet by comparing a configured sample rate with a current sample rate; upon determining that the current sample rate is less than the configured sample rate, exporting the queued packet from the network device; and upon determining that the current sample rate is greater than or equal to the configured sample rate, dropping the queued packet.