| In the era of continuous development and popularization of digital financialization,the amount of data shows explosive growth with the increase of business,and the concept of distributed is more and more popular.However,because most traditional banking core systems still use centralized mainframe systems,more and more bottlenecks are encountered in data expansion,system upgrade and data security,so it is imperative to migrate from the host centralized system to the distributed system.For large-scale banking system migration tasks,application service migration,distributed routing forwarding and database migration are important tasks in system migration.This paper focuses on application service migration,re-adapting routing rules before service forwarding,service reorganization and Research on the transaction consistency guarantee after optimization and service requests and propose some feasible migration solutions.The innovations are as follows.Firstly,a multi-level cluster distribution model of distributed system based on horizontal splitting of product types and vertical splitting of business functions is proposed.This model is suitable for the reconstruction or migration of complex large-scale business systems,which significantly improves the shortcomings of wide and uneven data distribution that affect system performance,and effectively exerts the development agility and flexibility of microservices through resource tilting.Secondly,a routing and forwarding strategy based on the combination of data distribution algorithm and load balancing algorithm is proposed.Due to the scattered data on the distributed system and the wide range of application deployment,when the application is migrated from the host system to the distributed system,the routing and addressing rules need to be re-adapted.According to factors such as data attributes and regional characteristics,combined with cluster splitting rules and Apply deployment rules to correctly forward requests to target servers by using consistent hash-based data distribution and load balancing algorithms.Finally,a scheme based on TCC mechanism to solve transaction consistency and based on pipeline tracking platform to ensure post-event consistency is proposed.Since system migration is a huge task,different service migration plans will be made at different stages,so services will be split,reorganized,and called across clusters,which may lead to transaction inconsistencies.In order to solve the transaction inconsistency problem caused by the reorganization service,this paper adopts the two-phase commit model of TCC,and designs the transaction flow tracking mode to register and detect the transaction status of the service,which solves the problem of transaction consistency to the greatest extent.In the system migration and reconstruction,the system performance has been greatly improved through the above improvement solutions.In addition,this paper also introduces new thinking and suggestions for the follow-up prospect of system distributed migration. |
PDF Full Text Request