Packet Re-marking for Queue Length Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current congestion control algorithms, such as TCP's slow-start mechanism, often lead to prolonged stabilization times and excessive packet loss when a new flow starts or restarts, as they lack accurate information about available capacity, causing buffer overflows and disruptions to other flows.
Innovation Solution
A method and apparatus for packet networks that use status tags to indicate queue length, allowing for explicit measurement and feedback of queue length, enabling senders to adjust their transmission rates more accurately and prevent congestion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If TCP uses slow-start algorithm to exponentially increase sending rate, then the flow can quickly probe for available capacity, but it causes excessive packet loss and long stabilization time
Solution Approach 1:
The patent applies preliminary action by having the sender transmit probe packets at the beginning of a flow to proactively discover available capacity before actual data transmission. This allows the system to gather congestion information in advance, avoiding the need for aggressive exponential rate increases that cause packet loss. The probe packets are sent with low priority to minimize disruption while still providing accurate capacity measurements.
Solution Approach 2:
The patent implements feedback mechanisms where the receiver sends congestion feedback packets to the sender based on queue length observations. This continuous feedback loop allows the sender to adjust its transmission rate dynamically without needing to aggressively probe for capacity. The feedback includes information about current queue length and recommended sending rates, enabling smooth convergence to optimal transmission rates without excessive packet loss.
2Productivity
If TCP doubles its congestion window every round trip during start-up, then it can rapidly increase sending rate, but it causes buffer overflow and disrupts other flows
Solution Approach 1:
The patent uses preliminary action by sending probe packets before the main data flow to measure available capacity in advance. This allows the sender to determine an appropriate initial congestion window size based on actual network conditions rather than arbitrarily doubling the window, preventing buffer overflow from the start.
Solution Approach 2:
The patent applies partial action by sending only a limited number of probe packets at low priority rather than immediately sending full-rate data. This partial probing provides sufficient capacity information without causing the excessive action of buffer overflow. The probe packets are intentionally kept to a minimum necessary for measurement while prioritizing other traffic.
3Loss of time
If a new flow starts with a large initial congestion window, then it can utilize available capacity faster, but it causes overshoot and excessive packet loss
Solution Approach 1:
The patent resolves this contradiction by performing preliminary capacity measurement using probe packets sent before the main data flow. This advance measurement provides accurate information about available capacity, allowing the sender to set an appropriate initial congestion window that is large enough to utilize capacity quickly but small enough to avoid overshoot and packet loss.
Solution Approach 2:
The patent replaces the mechanical doubling mechanism of traditional TCP with a measurement-based approach. Instead of mechanically doubling the congestion window regardless of actual conditions, the system substitutes this with probe packet measurements that provide real information about available capacity, enabling more reliable and efficient capacity utilization.
4Reliability
If TCP uses conservative initial window size, then it avoids buffer overflow, but it takes a long time to discover available capacity
Solution Approach 1:
The patent resolves this contradiction by performing preliminary capacity measurement using probe packets sent before the main data flow. This advance measurement allows the system to establish an optimal initial window size that is both safe and efficient, eliminating the need to choose between conservative and aggressive approaches. The probe packets provide the information needed to set an appropriately large initial window without causing buffer overflow.
Data Source
Figure 1
Figure 2~4
Figure 5
AI summary
A method and apparatus for changing a packet tag status value from a first value to a different value upon said packet arriving at a buffer if there already is a packet in the buffer queue having the same first status value. A sequence of tag status values (112) in packets received at an end node is used to determine the queue length of the buffer in the packet network. An end node can thereafter adapt its sending rate to the rate of the buffer.