A Perspective on Stable Wired Networks while Reducing Energy Use with Load Balance

The impact of the Internet and wired networks' growing energy consumption on costs and environment continues to be a major concern, while energy-efficient networks are attaining further interest in the research community. We examined the problem of reducing power consumption (PC), while accomplishing acceptable load balance by altering link weights to allow network traffic to be serviced by only partial of the network. Those unused network devices and links can therefore be turned off to reduce the network energy consumption. This study created a new approach that uses live data from wired networks to perform Multi-Objective Optimization (MOO) utilizing a Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimize the PC and Maximum Link Utilization (MLU) objectives. However, changing link weights frequently may cause network instability due to the network traffic being routed from different shortest paths back and forth, which is a situation to be avoided. To address this issue, we investigated how introducing delays between link weight adjustments can reduce network oscillation or instability without increasing network power consumption or affecting traffic imbalances. We introduced the sequential_delay and wait_interval delay parameters, which work in tandem with link weight settings. We showed that average_δ
instability
, the metric we used to measure the network instability, was reduced from 145.5% down to 14.67% when using a value of 1sec for both sequential_delay and wait_interval. The hybrid_stable approach, which combines the hybrid approach and applies the delay parameters on link weight updating, not only improved the network instability, but also reduced more PC and MLU. The simulation results show that PC was reduced by 43.75%, while it reduced MLU by 43.24%.