Packet Core Load Balancing for Latency and Power-Aware Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional networking systems face inefficiencies in distributing network traffic due to unequal processing time requirements and core performance, leading to underutilization, overutilization, increased power consumption, and latency.
Innovation Solution
A dynamic load balancing system using a load balancing unit that employs machine learning algorithms to determine the optimal distribution of workflow packets among packet processing cores based on utilization, power states, physical distance, and packet priority, utilizing reinforcement learning to iteratively improve scheduling policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional static load balancing is used to distribute network traffic, then system simplicity is maintained, but core utilization becomes unequal leading to underutilization and overutilization of processing cores
Solution Approach 1:
The patent implements dynamic load balancing that continuously monitors and adjusts packet distribution based on real-time core utilization metrics. The system transitions from static to dynamic routing decisions, adapting to changing workload conditions to optimize core utilization without requiring overly complex infrastructure
Solution Approach 2:
The system incorporates feedback mechanisms where utilization data from multiple packet processing cores is collected and used to inform subsequent routing decisions. This closed-loop approach allows the load balancer to learn from system performance and continuously improve packet distribution, resolving the contradiction between simplicity and optimization
2Speed
If packets are routed without considering core performance characteristics, then routing simplicity is maintained, but processing latency increases due to mismatched packet-core assignments
Solution Approach 1:
The patent applies local quality by matching specific packet characteristics with suitable core capabilities. Different cores are assigned packets based on their individual performance profiles, creating optimized local processing paths that reduce latency while maintaining overall system simplicity through rule-based matching
3Use of energy by moving object
If load balancing does not account for power states, then system operation simplicity is maintained, but power consumption increases due to inefficient core selection
Solution Approach 1:
The system changes the selection parameters for core assignment to include power state information. By incorporating power consumption metrics into the routing decision framework, the system optimizes energy usage while maintaining manageable complexity through integrated parameter evaluation
4Productivity
If reinforcement learning is implemented to optimize packet routing, then system performance and adaptability are improved, but computational overhead and processing time increase
Solution Approach 1:
The system performs preliminary actions by pre-training the reinforcement learning model offline to establish optimal routing policies. During runtime, the pre-trained model makes rapid routing decisions without requiring extensive real-time computation, thus improving performance while minimizing processing time overhead
Data Source
AI summary
Methods, systems, and computer program products for selecting packing processing cores are provided. An example system includes a plurality of packet processing cores and a load balancing unit communicatively connected to the plurality of packet processing cores. The load balancing unit is configured to receive a workflow packet including packet description data indicative of at least a packet structure and a packet priority and receive, from the plurality of packet processing cores, state data indicative of at least a utilization state and an operating state of each of the respective packet processing cores. The load balancing unit determines a selected packet processing core from amongst the plurality of packet processing cores based on the state data of the packet processing core and the packet description data of the workflow packet and transmits the workflow packet to the selected packet processing core.