Flow Selector Classifying Elephant Mice Flows for Network Resource Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data center networks, large-sized flows (elephant flows) can cause congestion and delay, as they transfer more data over a longer duration, but existing solutions often detect these flows only after they have completed, making it ineffective in preventing current delays.
Innovation Solution
Implementing a flow selector that classifies flows into elephant and mice flows using machine-learning algorithms, allowing for real-time network resource selection based on flow characteristics, and assigning weights to network resources to optimize traffic distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If equal cost multi-path routing (ECMP) is used to distribute flows equally across multiple paths, then network resource utilization is improved, but congestion and delay occur on paths carrying larger elephant flows
Solution Approach 1:
The patent applies local quality by differentiating flow handling based on flow type. Elephant flows (large data transfers) are routed through dedicated paths with higher bandwidth capacity, while mice flows (small data transfers) use standard ECMP paths. This localized optimization ensures that each flow type receives appropriate network resources, preventing congestion on paths carrying elephant flows while maintaining overall network utilization.
Solution Approach 2:
The patent changes the routing parameter from uniform path selection to flow-type-aware path selection. By classifying flows into elephant and mice categories based on characteristics such as data size and duration, the system dynamically adjusts routing decisions. Elephant flows are assigned to paths with sufficient capacity to handle large transfers, while mice flows continue to use load-balanced paths, thereby optimizing both resource utilization and flow completion times.
2Measurement precision
If flow detection is performed after flows complete, then accurate identification of elephant flows is achieved, but real-time prevention of network delay is not possible
Solution Approach 1:
The patent implements preliminary action by classifying flows into elephant and mice types before they complete their transmission. Using flow classification based on characteristics observed during flow establishment and early transmission phases, the system proactively routes elephant flows through appropriate paths before congestion can occur. This preemptive classification and routing prevents delay rather than detecting it after completion.
Solution Approach 2:
The system employs feedback mechanisms by continuously monitoring flow characteristics and adjusting routing decisions in real-time. Flow classification information feeds back into the routing logic, enabling dynamic path selection that adapts to current network conditions and flow types, thereby preventing congestion before it impacts performance.
Data Source
AI summary
In some embodiments, a method receives a set of packets for a flow and determines a set of features for the flow from the set of packets. A classification of an elephant flow or a mice flow is selected based on the set of features. The classification is selected before assigning the flow to a network resource in a plurality of network resources. The method assigns the flow to a network resource in the plurality of network resources based on the classification for the flow and a set of classifications for flows currently assigned to the plurality of network resources. Then, the method sends the set of packets for the flow using the assigned network resource.