Design and Implementation of Live Cache Repair Without Process Restart for Authentication Workloads

Abstract

Authentication systems are important components of distributed applications in modern days, as an enduring availability and responsiveness with speed are necessary. In this paper, downtime refers to authentication request failures (5xx errors) or application process restart during repair. These mechanisms in most cases make use of caching mechanisms to enhance the rate of authentication; but when the caches are corrupted or contain stale elements as well as failures to maintain consistent state on the data, then it impacts adversely on the reliability of the services. The conventional recovery methods usually entail starting up the application again or hand repairing cache, which leads to service disruption of the service and low availability. The study presents a Python-implemented Operator-Initiated Live Cache Repair Framework, which allows real-time identification, authentication, and fixation of authentication cache records without terminating running processes. The suggested structure includes a monitoring screen of the cache, the authentication engine, a repair engine, operator interface, and a log. Synthetic authentication workloads and controlled cache fault injection are used to create a prototype system. Up to now, experimental assessment has shown that the framework presented decreases recovery time, better authentication success rate, enhances the availability of the system, and sustains adequate resource usage when repairing the system. Results show that runtime cache repair is an efficient technique used in enhancing fault tolerance and continuity of the authentication services.