Receive Window Control for Incast Congestion in Data Centers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data centers, competition among elephant flows leads to incast congestion, where multiple transmission-side end nodes transmit data parallelly to a reception-side end node, causing throughput inefficiencies due to the competition among these high-volume, long-lifetime flows.
Innovation Solution
The system implements an RWIN control unit in each reception-side end node, which monitors the congestion window, measures round trip time, calculates minimum throughput, and adjusts the receive window size to optimize throughput by setting a setting window size based on past and calculated values, and a controller calculates and transmits an average round trip time to manage competing elephant flows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If multiple transmission-side end nodes transmit data parallelly to a reception-side end node, then data transfer speed increases, but incast congestion occurs causing throughput inefficiency
Solution Approach 1:
The receive window size is dynamically adjusted based on monitored congestion window increases and measured round trip times. The RWIN control unit continuously adapts the receive window size to match current network conditions, transforming the static window size into a dynamic parameter that responds to changing traffic patterns and congestion levels.
Solution Approach 2:
The system implements feedback mechanisms where the reception-side end node monitors congestion window increases, measures round trip times, and uses this information to calculate and adjust the receive window size. This closed-loop feedback allows the system to automatically respond to network conditions and optimize throughput.
2Productivity
If the receive window size is increased to improve throughput, then data transfer efficiency improves, but incast congestion worsens
Solution Approach 1:
The RWIN control unit continuously monitors congestion window increases and round trip times, using this feedback to dynamically adjust the receive window size. This feedback mechanism allows the system to increase throughput while automatically responding to congestion conditions by adjusting the window size appropriately.
Solution Approach 2:
The system changes the receive window size parameter based on monitored network conditions. By adjusting this critical parameter dynamically rather than using a fixed value, the system can optimize throughput under normal conditions while preventing congestion when network conditions deteriorate.
3Object-generated harmful factors
If the receive window size is decreased to reduce incast congestion, then congestion is suppressed, but throughput decreases
Solution Approach 1:
The receive window size transitions from a static parameter to a dynamic one that automatically adjusts based on monitored congestion window increases and round trip times. This dynamic adjustment allows the system to maintain high throughput when possible while suppressing congestion when necessary.
Data Source
AI summary
A system includes: a transmitting device, switch devices, receiving devices, and a control device. Each of the receiving devices carries out a process including monitoring increase in size of a congestion window which the transmitting device includes, measuring a round trip time, calculating a minimum throughput in one or more target flows whose volume is not smaller than a given size among flows whose packets are received when the increase in the size of the congestion window becomes a steady state, creating a setting window size that is a receive window size for setting on a basis of an already-set window size that is size of a receive window set in a past response packet and a calculated window size that is size of a receive window calculated from the minimum throughput, and transmitting a response packet in which the setting window size is set to the transmitting device.