Machine Introspection for Elephant Flow Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Elephant flows in data networks cause non-trivial queuing delays and affect latency-sensitive mouse flows due to buffer filling, and existing adaptive routing techniques are ineffective for bursty mouse flows.
Innovation Solution
A system detects elephant flows by examining machine operations, using machine introspection to identify new data flows and track data amounts, and treats them differently from mouse flows through separate queuing or routing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If network buffers are used to handle high-volume elephant flows, then throughput for elephant flows is improved, but queuing delay increases for mouse flows sharing the same buffers
Solution Approach 1:
The patent segments network traffic into two categories: elephant flows and mouse flows. By classifying flows based on their characteristics (volume, duration, burstiness), the system applies different handling strategies to each segment, allowing high-volume elephant flows to receive dedicated buffer resources while mouse flows get prioritized handling to minimize queuing delay
Solution Approach 2:
The patent applies different quality characteristics to different parts of the traffic flow. Elephant flows receive high throughput optimization with larger buffer allocations, while mouse flows receive low-latency optimization with priority queuing and smaller buffer allocations. This local differentiation resolves the contradiction by tailoring buffer handling to the specific needs of each flow type
2Productivity
If adaptive routing techniques are applied to mouse flows, then routing efficiency is improved, but effectiveness decreases due to the bursty nature of mouse flows
Solution Approach 1:
The patent introduces dynamic flow classification that adapts to the actual behavior of traffic flows. By continuously monitoring flow characteristics such as packet arrival patterns, duration, and volume, the system dynamically adjusts routing decisions. This dynamic approach allows mouse flows to be identified and handled differently from elephant flows, making routing effective despite the bursty nature of mouse traffic
Data Source
AI summary
Some embodiments provide a system that detects whether a data flow is an elephant flow; and if so, the system treats it differently than a mouse flow. The system of some embodiments detects an elephant flow by examining, among other items, the operations of a machine. In detecting, the system identifies an initiation of a new data flow associated with the machine. The new data flow can be an outbound data flow or an inbound data flow. The system then determines, based on the amount of data being sent or received, if the data flow is an elephant flow. The system of some embodiments identifies the initiation of a new data flow by intercepting a socket call or request to transfer a file.