Data Packet Network Congestion Control via Probe Packets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data packet networks face challenges in determining the appropriate transmission rate due to outdated congestion feedback, leading to overshoot and significant packet loss, especially as network capacities increase, causing instability and damage to other data flows.
Innovation Solution
Implementing a new class of service for data packets that are not queued if a buffer is occupied, allowing these packets to be discarded at intermediate nodes, thereby providing immediate feedback to adjust transmission rates and reducing packet loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If closed-loop congestion control algorithms are used to determine transmission rate, then the sender node can adjust transmission rate based on congestion feedback, but the congestion feedback becomes outdated quickly causing inaccurate rate prediction and overshoot
Solution Approach 1:
The patent introduces probe packets that are sent beforehand to measure network capacity before actual data transmission begins. This preliminary measurement allows the sender to determine the appropriate transmission rate in advance, avoiding the lag inherent in closed-loop feedback mechanisms. The probe packets elicit capacity indicators from intermediate nodes, providing timely congestion information without waiting for data packet acknowledgments.
2Productivity
If TCP Slow-Start exponential growth phase is used to increase transmission rate, then the sender can quickly utilize available bandwidth, but it causes significant packet loss when exceeding network capacity
Solution Approach 1:
The patent performs capacity measurement using probe packets before initiating the exponential growth phase of data transmission. This preliminary action provides an accurate baseline of network capacity, allowing the sender to start data transmission at a sustainable rate rather than aggressively doubling the congestion window, thereby avoiding overshoot and packet loss while still achieving efficient bandwidth utilization.
Solution Approach 2:
The patent uses probe packets as disposable objects to measure network capacity. These probe packets are sent temporarily to gather capacity information and then discarded, allowing the system to obtain accurate network state information without committing to long-term transmission decisions based on potentially outdated feedback. The probe packets serve their purpose and are replaced as needed to maintain accurate capacity knowledge.
3Reliability
If intermediate nodes queue all incoming packets, then packet delivery is maintained during congestion, but transmission rate information is delayed and overshoot occurs
Solution Approach 1:
The patent segments packets into two types: probe packets and data packets. Probe packets are treated differently from data packets in terms of queuing behavior. When the buffer is full, data packets are dropped while probe packets can still be processed to elicit capacity indicators. This segmentation allows the system to maintain reliability for data packets while obtaining timely capacity information through probe packets, resolving the contradiction between queueing for reliability and speed for feedback.
Data Source
AI summary
The invention relates to a method of controlling packets in a data packet network, the method comprising the steps of: a source node sending a first set of packets to a receiver node over the data packet network via an intermediate node, wherein the first set of packets are marked as being of a class of service preventing them from being forwarded to the receiver node if a packet queue exists at an intermediate node; the source node receiving an acknowledgement of a received packet of the first set of packets from the receiver node; and the source node sending a second set of packets to the receiver node over the data packet network.