Bandwidth Estimation via Packet Skipping in OWD Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bandwidth estimation techniques, such as PathCos++ and SLDRT, face significant errors due to capacity fluctuations and One-Way Delay (OWD) noise, especially in scenarios like cold start and bursting conditions, which affect the accuracy of available bandwidth estimation in network paths.
Innovation Solution
The proposed solution involves using Inter-Arrival Time (IAT) fitting techniques to skip initial packets from the probe train that are affected by capacity fluctuations, allowing for the selection of a packet pair that better represents the long-term average packet rate, thereby reducing estimation errors and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional bandwidth estimation techniques (PathCos++, SLDRT) are used, then bandwidth measurement can be performed, but significant errors occur due to capacity fluctuations and OWD noise
Solution Approach 1:
The patent applies preliminary action by pre-calculating expected OWD values based on packet pair index and sending parameters before actual packet transmission. This allows the system to predict and compensate for OWD noise and capacity fluctuations in advance, rather than reacting to them after they affect the measurement. The expected OWD is computed using a formula that considers packet size, link capacity, and queue depth, enabling proactive correction of measurement errors.
Solution Approach 2:
The patent implements feedback by continuously comparing actual OWD measurements with expected OWD values and using this difference to adjust bandwidth estimation. The system calculates OWD noise as the difference between actual and expected delays, then uses this feedback to correct the bandwidth estimation formula, thereby compensating for capacity fluctuations and improving measurement accuracy dynamically.
2Productivity
If probe packets are sent through the network path, then bandwidth can be measured, but capacity fluctuations during packet transmission cause large errors in estimation
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the bandwidth estimation calculation based on varying network conditions. Instead of using a fixed estimation formula, the system modifies the effective bandwidth calculation by subtracting estimated OWD noise from the raw measurement. This adaptive parameter adjustment allows accurate bandwidth estimation even when network capacity fluctuates during packet transmission.
Solution Approach 2:
The patent replaces direct mechanical measurement of bandwidth with a computational model that calculates expected OWD based on packet characteristics and network parameters. Instead of relying solely on raw packet transmission timing, the system substitutes a mathematical model that accounts for queueing theory and link characteristics, thereby eliminating errors caused by capacity fluctuations during measurement.
3Measurement precision
If direct measurement on network devices is used, then accurate bandwidth data can be obtained, but it is not possible when devices are in different administrative domains or when characteristics are hidden by tunneling
Solution Approach 1:
The patent introduces probe packets as intermediaries that traverse the network path between endpoints in different administrative domains. These probe packets carry timing information and are used to indirectly measure bandwidth without requiring direct access to network device internals. The measurement is performed end-to-end using the probe packets as mediators, enabling bandwidth estimation across distributed networks where direct measurement is impossible.
Solution Approach 2:
The patent segments the bandwidth measurement problem into separate components: probe packet transmission, timing measurement at endpoints, and computational estimation. This segmentation allows the measurement to be performed independently at each endpoint using only local resources and probe packet data, eliminating the need for direct access to network devices and enabling deployment in multi-domain environments with tunneling.
Data Source
AI summary
Techniques and architectures for measuring available bandwidth. A train of probe packets is received from a remote electronic device. A received time for a first subset of the train of probe packets is measured. A change in capacity of the network path is determined based on the measured received time for the first subset of packets. Packets from the train of probe packets prior to the detected change in capacity of the network path are excluded to identify a second subset of packets. An estimated available bandwidth is computed based on the second subset of packets from the train of probe packets.