Energy-Aware Weighted ECMP Routing for Data Center Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large data centers face increasing energy usage as they grow, with conventional methods unable to dynamically improve energy efficiency in response to changing configurations and workloads.
Innovation Solution
Classifying packet flows across different paths in a computer network based on an energy cost metric, such as the green-BitCost metric, to load balance packets and optimize energy usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional data centers source energy from renewable sources, then energy efficiency is improved, but the data centers cannot dynamically increase energy efficiency in response to changing configurations and workloads
Solution Approach 1:
The patent implements dynamic path selection by modifying ECMP routing to consider real-time energy costs of different network paths. The system continuously monitors and recalculates energy costs based on current workload and configuration, allowing the network to dynamically adapt its routing decisions to minimize energy consumption while responding to changing conditions.
Solution Approach 2:
The patent introduces energy cost as a new parameter in routing decisions, modifying the traditional ECMP cost calculation. By incorporating energy cost metrics into the path selection algorithm, the system can dynamically adjust routing based on energy consumption patterns, transforming static routing into energy-aware dynamic routing.
2Productivity
If data centers grow larger to meet increasing demand, then service capacity is improved, but energy usage increases significantly
Solution Approach 1:
The patent segments the network into multiple paths with different energy cost characteristics. By dividing the network traffic into different path segments and selecting paths based on their energy costs, the system can handle increased service capacity while minimizing the energy consumption of each transmission segment.
Solution Approach 2:
The patent applies local quality optimization by assigning different energy cost weights to different network paths based on their specific characteristics (distance, hop count, link capacity). This allows the system to optimize energy consumption locally for each path while maintaining overall network capacity for handling increased service demands.
3Reliability
If multiple network paths are used for load balancing, then network reliability is improved, but energy consumption increases due to traffic distribution across more links
Solution Approach 1:
The patent dynamically adjusts path selection for load balancing by considering energy costs. Instead of uniformly distributing traffic across all available paths, the system dynamically selects paths based on their energy cost metrics, maintaining load balancing benefits while minimizing energy consumption through intelligent path selection.
Solution Approach 2:
The patent implements feedback mechanisms by monitoring energy consumption patterns and using this information to adjust routing decisions. The system continuously evaluates the energy cost of different paths and feeds this information back into the routing algorithm, allowing it to optimize the balance between reliability and energy consumption based on real-time conditions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques are described for classifying packet flows to different paths of a computer network based at least in part on an energy cost of each of the different paths. In one example, a network device determines a weighted equal-cost multipath (wECMP) cost of each of different paths over which to forward packets. The different paths may include, for example, different interfaces of a network device, different links to which the network device is connected, or different links of an aggregated bundle of Ethernet links. The network device determines an energy cost of each of the paths. The network device modifies the wECMP cost of each path based at least in part on the energy cost of the path to obtain a modified wECMP cost. The network device load balances the packets over the paths in accordance with the modified wECMP cost.