基于TCN-GRU的Handle标识解析系统负载均衡算法

Load Balancing Algorithm for Handle Identification Resolution System Based on TCN-GRU

数联网技术是解决数据管理困难及数据孤岛等问题的关键。近年来,随着数联网技术的飞速发展,Handle系统作为数联网标识解析领域中最具有代表性的系统得以快速落地。然而,Handle标识注册量和解析请求量的快速增长使得Handle系统缓存/递归解析服务节点的解析请求压力加大,且难以保证标识解析请求的服务质量。针对上述问题,提出一种基于TCN-GRU的缓存/递归解析服务节点的负载均衡算法,旨在优化缓存/递归解析服务节点的标识解析效率。首先,将用户的解析请求合理地分配到缓存/递归解析服务节点的服务器集群当中,以提升服务器集群整体的工作效率和吞吐量;其次,基于Handle系统中节点的层级架构,考虑到循环神经网络中普遍存在的梯度弥散问题,引入TCN时序神经网络,有助于提取服务器集群负载的时序性信息。相较于传统的负载均衡算法,所提算法根据服务器负载状况动态调整负载转发策略,服务器平均响应时间缩短50%左右,吞吐率提高40%左右,请求失败次数大幅减少。

Internet of Data(IoD)technology is key to solving challenges related to data management difficulties and data silos.In recent years,the IoD technology has witnessed rapid development,and the Handle system has emerged as a prominent system for IoD identification resolution.However,the significant increase in Handle identity registrations and resolution requests has intensified the pressure on Handle system cache/recursive resolution service nodes,making it difficult to ensure the quality of identity resolution services.To address these challenges,this paper proposes a load balancing algorithm based on TCN-GRU cache/recursive resolution service nodes,aiming to optimize the identity resolution efficiency of these nodes.Firstly,the algorithm intelligently distributes user resolution requests across the server cluster of cache/recursive resolution service nodes,thus enhancing the efficiency and throughput of the entire server cluster.Secondly,leveraging the hierarchical architecture of nodes in the Handle system and considering the common gradient diffusion problem in recurrent neural networks,the TCN time-series neural network is introduced to extract temporal information from the server cluster load.In comparison to traditional load balancing algorithms,the proposed approach dynamically adjusts the load forwarding strategy based on the server load status,resulting in approximately a 50%reduction in average server response time,about 40%increase in throughput rate,and a significant decrease in request failures.