Neural Network Flow Table for Data Center Traffic Segregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data center networks face bottlenecks due to the inefficiencies in electrical packet switching, particularly when handling large amounts of data, which slows down communication and causes congestion, as electrical switches must determine the destination for each packet individually, leading to increased latency and reduced performance.
Innovation Solution
The integration of a neural network within network devices to analyze packet flows, determine characteristics such as 'elephant flows,' and use this information to dynamically route packets through either electrical or optical switches, optimizing the flow by segregating high-bandwidth traffic through optical circuit switches and low-bandwidth traffic through electrical packet switches, thereby reducing congestion and enhancing network efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If electrical packet switching is used to handle all traffic, then flexibility and adaptability are maintained, but network congestion and latency increase due to per-packet processing overhead
Solution Approach 1:
The patent segments traffic into different types (elephant flows and mouse flows) and routes them through different switching paths. Elephant flows are directed through optical circuit switches for high-speed transmission, while mouse flows continue to be handled by electrical packet switches. This segmentation resolves the contradiction by allowing high-volume traffic to bypass the processing overhead of electrical switches while maintaining flexibility for other traffic types.
Solution Approach 2:
The patent introduces an intermediary classification mechanism that identifies elephant flows and redirects them to optical circuit switches. This intermediary layer acts as a mediator between the electrical packet switch and optical circuit switch, determining which path each flow should take based on its characteristics, thereby improving overall network throughput without sacrificing adaptability.
2Speed
If optical circuit switching is used for high-bandwidth traffic, then transmission speed and bandwidth are improved, but device complexity increases due to the need for flow classification and dynamic reconfiguration
Solution Approach 1:
The patent applies preliminary action by classifying flows and making routing decisions before traffic enters the optical circuit switch. The electrical packet switch or a dedicated classification mechanism identifies elephant flows and pre-configures the optical circuit switch accordingly, so that when high-bandwidth traffic arrives, the path is already established and ready for high-speed transmission without requiring complex real-time reconfiguration.
Solution Approach 2:
The system implements self-service by using the electrical packet switch to automatically identify and classify elephant flows based on their traffic characteristics, then redirecting them to the optical circuit switch. This self-classification mechanism reduces the need for external control and manual configuration, simplifying the overall system while enabling high-speed optical transmission for appropriate traffic.
3Adaptability or versatility
If dynamic reconfiguration of optical circuit elements is performed, then network adaptability and bottleneck prevention are improved, but processing time and control overhead increase
Solution Approach 1:
The patent performs preliminary classification and routing decisions for elephant flows before they enter the optical circuit switching fabric. By identifying these high-volume flows early and pre-configuring the optical paths, the system minimizes the need for frequent dynamic reconfiguration during active transmission, thereby reducing reconfiguration time and processing overhead while maintaining network adaptability.
Data Source
AI summary
A flow of packets is communicated through a data center. The data center includes multiple racks, where each rack includes multiple network devices. A group of packets of the flow is received onto an integrated circuit located in one of the network devices. The integrated circuit includes a neural network and a flow table. The neural network analyzes the group of packets and in response determines if it is likely that the flow has a particular characteristic. The neural network outputs a neural network output value that indicates if it is likely that the flow has a particular characteristic. The neural network output value, or a value derived from it, is included in a flow entry in the flow table on the integrated circuit. Packets of the flow subsequently received onto the integrated circuit are routed or otherwise processed according to the flow entry associated with the flow.