Virtual Converged Enhanced Ethernet Flow Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Converged Enhanced Ethernet (CEE) networks, especially over extended distance links, there are significant delays due to credit-based flow control, leading to degraded performance and limited useful link distance, as each I/O request waits for flow control replies, causing inefficiencies and performance limitations.
Innovation Solution
A method that identifies links with high response times and transmits data packets without waiting for acknowledgments (ACKs), using a large credit number to maintain full bandwidth, allowing for continuous data streaming and re-transmitting invalid packets upon link failure, thereby optimizing bandwidth utilization and performance across longer distances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If credit-based flow control is used in CEE networks, then data transmission reliability is maintained, but transmission speed and bandwidth utilization deteriorate due to delays in waiting for acknowledgments
Solution Approach 1:
The patent pre-calculates and pre-allocates a large credit number based on historical ACK reception patterns and link characteristics before actual data transmission begins. This preliminary action allows the transmitter to send multiple data packets without waiting for individual ACKs, thereby maintaining reliability through pre-planned credit allocation while significantly improving transmission speed by eliminating wait states.
Solution Approach 2:
The patent uses excessive action by allocating more credits than traditionally required (e.g., 256 credits instead of typical small numbers). This excessive credit allocation enables the transmitter to maintain continuous data flow over extended periods without ACK interruptions, effectively decoupling transmission speed from ACK latency and resolving the contradiction between reliability and speed.
2Measurement precision
If credit-based flow control waits for ACKs before transmitting each data packet, then transmission accuracy is ensured, but bandwidth utilization and productivity deteriorate
Solution Approach 1:
The system performs preliminary analysis of ACK reception patterns and link characteristics to pre-determine an optimal large credit number. This preliminary action enables accurate transmission by pre-calculating the credit allocation needed to maintain reliability, while simultaneously improving productivity by allowing continuous data transmission without repeated ACK wait cycles.
Solution Approach 2:
The patent fundamentally changes the credit parameter from traditional small values to large pre-calculated values (e.g., 256 credits). This parameter change transforms the transmission mode from ACK-dependent stop-and-wait to continuous flow, maintaining transmission accuracy through careful parameter selection while dramatically improving bandwidth utilization and overall productivity.
3Device complexity
If traditional flow control is used over extended distance links, then protocol simplicity is maintained, but useful link distance and adaptability deteriorate due to performance degradation
Solution Approach 1:
The patent extends adaptability to long distances by changing the credit parameter from small traditional values to large pre-calculated values. This parameter change allows the simple credit-based flow control protocol to accommodate extended distance links with high latency, as the large credit number absorbs the delayed ACK return time, enabling the same simple protocol to work effectively over both short and long distances.
Solution Approach 2:
The system performs preliminary characterization of the link (distance, latency, bandwidth) to pre-calculate the appropriate large credit number before deployment. This preliminary action enables the simple protocol to adapt to different link distances without increasing protocol complexity, as the adaptation is achieved through pre-computed parameters rather than complex runtime negotiations.
Data Source
AI summary
Maintaining bandwidth in links between servers and storage arrays comprising a device. The device establishes the links. The device identifies a first link from the links. The first link has a high response time. The device transmits a plurality of data packets on the first link. Each data packet is associated with a corresponding acknowledgment (ACK). The transmission is performed without waiting for the corresponding ACK to be received. The device tracks the ACK received in response to each of the transmitted data packets. The device detects a failure of the first link. In response to the detection, the device identifies invalid data packets. The invalid data packets comprise data packets transmitted on the first link after the detected failure.