Flow-Level Congestion Control for Microburst and Deadlock Relief
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current congestion control techniques in networks suffer from inefficiencies such as microbursts, network-wide deadlocks, and slow, imprecise congestion notifications, leading to packet drops and reduced network performance.
Innovation Solution
A system using a flow-level transmit mechanism with a source and receive SFA that generates and processes congestion notification packets (CNP) to identify and throttle specific transmit queues causing congestion, providing fast and precise congestion control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If segment offloading is applied to reduce processing overhead, then CPU processing overhead is reduced, but microbursts are created that overflow packet buffers and cause packet drops
Solution Approach 1:
The patent segments the flow control mechanism into per-flow pause frames rather than per-queue or per-COS control. Each flow receives individual pause frames based on its own buffer status, allowing fine-grained control that prevents microbursts while maintaining high throughput. This segmentation at the flow level resolves the contradiction by enabling precise control without overwhelming the system.
Solution Approach 2:
The patent implements feedback mechanisms where receivers send pause frames to senders based on real-time buffer status detection. This feedback loop allows the system to dynamically adjust transmission rates per flow, preventing buffer overflow and packet drops while maintaining efficient processing. The feedback mechanism enables the system to respond to congestion conditions without creating harmful microbursts.
2Reliability
If priority-based flow control is used to relieve congestion in a specific CoS, then that CoS congestion is relieved, but all flows of the same CoS group pause excessively causing network-wide deadlocks
Solution Approach 1:
The patent segments flow control at the individual flow level rather than grouping by CoS. Each flow is independently monitored and controlled based on its own buffer status, eliminating the excessive pausing that occurs when entire CoS groups are paused. This segmentation allows selective congestion relief without causing network-wide deadlocks, resolving the contradiction between effective congestion relief and maintaining network throughput.
Solution Approach 2:
The patent applies local quality control by treating each flow differently based on its specific buffer status and congestion conditions. Instead of uniform CoS-based control, the system applies tailored pause frames to individual flows that are actually congested, while allowing other flows to continue transmitting. This local化的 control approach resolves the contradiction by providing precise congestion relief without unnecessarily pausing all flows in a CoS group.
3Loss of information
If ECN is used to notify senders of congestion, then congestion notification is provided, but responses are slow and imprecise because it takes a long time for the sender to receive notification and throttle the flow
Solution Approach 1:
The patent applies preliminary action by having receivers prepare and send pause frames immediately when buffer thresholds are reached, before significant congestion develops. This proactive approach reduces the response time compared to reactive ECN mechanisms that wait for congestion to manifest. The receiver proactively notifies the sender to pause transmission, eliminating the delay inherent in ECN's mark-and-notify approach.
Solution Approach 2:
The patent implements a direct feedback mechanism where receivers send pause frames to senders based on real-time buffer status. This feedback loop operates at the flow level and provides immediate notification of congestion conditions, eliminating the slowness and imprecision of ECN. The feedback mechanism allows senders to throttle their flows rapidly in response to actual buffer status, resolving the contradiction between notification accuracy and response time.
4Productivity
If traffic continues flowing at full rate during congestion, then maximum throughput is maintained, but the receiver is overwhelmed and packet drops occur
Solution Approach 1:
The patent uses feedback mechanisms where receivers continuously monitor buffer status and send pause frames to senders when thresholds are reached. This feedback loop enables dynamic adjustment of transmission rates to match receiver capacity, preventing buffer overflow and packet drops while maintaining optimal throughput. The feedback mechanism allows the system to operate close to maximum capacity without overwhelming the receiver.
Solution Approach 2:
The patent implements dynamic flow control where transmission rates are continuously adjusted based on real-time buffer status. Rather than maintaining static full-rate transmission, the system dynamically modulates flow rates per individual flow based on actual congestion conditions. This dynamic approach allows the system to maintain high throughput when possible while preventing packet drops when congestion occurs, resolving the contradiction between throughput and reliability.
Data Source
AI summary
A system for congestion control using a flow level transmit mechanism is disclosed. In some embodiments, the system comprises a source SFA and a receive SFA. The source SFA is configured to detect and classify a congestion notification packet (CNP) generated based on congestion in a network; select a receive block from a plurality of receive blocks based on the CNP; forward the CNP to a dedicated congestion notification queue of the receive block; identify a transmit queue from a plurality of transmit blocks based on processing the congestion notification queue, wherein the transmit queue originated a particular transmit flow causing the congestion; and stop the transmit queue.