Modern distributed systems store thousands of gigabytes of information in persistent relational database management systems. It is the core storage component for microservice-based architectures. Due to the peculiarities of internal storage mechanisms, the total query processing time may exceed the service level agreement value. A common solution is to add a caching layer between the database management system and the service. However, maintaining the cache in a consistent state across all service replicas is a challenging task. We believe we can design a caching system with a consensus algorithm to meet modern service level agreement requirements. Firstly, we outline the role of caching in a distributed systems context. Secondly, we clarify our consistency model choice. Thirdly, we analyze existing distributed systems and their consistency guarantees to outline the absence of a solution that can fit our requirements. Finally, we develop the architecture of the caching system.