Fine-grained single-flow delay detection method based on reference delay insertion structure

Abstract

The invention provides a fine-grained single-flow delay detection method based on a reference delay insertion structure, and the method comprises the following steps of setting a reference packet generator at a transmission end; setting a delay estimator at a receiving end; and periodically injecting a reference packet with a timestamp from the transmission end into a data stream by utilizing the reference packet generator, delaying the reference packet in the transmission process, receiving the reference packet delay sample by utilizing the receiving end, and collecting the reference packet delay sample and performing the delay detection by utilizing the delay estimator. By setting the reference packet, as long as the transmission time of the reference packet is close enough to the transmission time of the data packet, the delay property of the data packet can be deduced after observing the reference packet, and thus the delay estimation error can be reduced. In addition, compared with the scheme that a great amount of data packet timestamps are required to collect from multiple terminals, the detection result can be directly acquired on the receiving end, and the timestamps of all conventional data packets on the transmission end are not required, so that the architecture is convenient to implement, the cost performance is high, and wide popularization and application can be favored.

Description

technical field [0001] The invention relates to the technical field of computer multimedia, in particular to a fine-grained single-stream-level delay detection method based on a reference delay insertion structure. Background technique [0002] At present, the Internet is gradually being applied to many new fields, such as algorithmic trading and high-performance computing. These application fields have very strict requirements on latency, reaching the microsecond level. However, today's network operators lack fine-grained detection tools to locate and correct network anomalies and latency spikes that cause SLA violations. Recently, some researchers have proposed a scalable delay detection scheme called LDA, but it can only perform delay detection on aggregated flows, and because different flows will show very obvious differences even if they are transmitted on the same link, Therefore, we need delay detection of individual streams when troubleshooting special faults at the...

AI Technical Summary

Problems solved by technology

However, today's network operators lack fine-grained detection tools to locate and correct network anomalies and latency spikes that cause SLA violations

Recently, some researchers have proposed a scalable delay detection scheme called LDA, but it can only perform delay detection on aggregated flows, and because different flows will show very obvious differences even if they are transmitted on the same link, Therefore, when troubleshooting special faults at the application level, we need delay detection of a single stream

There is also a delay detection scheme based on active probe detection technology. This detection technology is mainly used in end-to-end systems and cannot detect router delay problems. In addition, expensive high-precision detection equipment cannot be widely used in today's networks.

Images