This article is devoted to the study of problematic issues due to the growing scale and requirements for modern high-load distributed systems. The relevance of the work is ensured by the fact that an important component of each such system is a database. The paper highlights the main problems associated with the use of relational databases in many high-load distributed systems. The main focus is on the study of such properties as data consistency, availability, and stability of the system. Basic information about the architecture and purpose of non-relational databases with a wide column, databases of key-value type, and document-oriented databases is provided. The advantages and disadvantages of non-relational databases of different types are shown, which are manifested in solving different problems depending on the purpose and features of the system. The choice of non-relational databases of different types for comparative analysis is substantiated. Databases such as Cassandra, Redis, and Mongo, which have long been used in high-load distributed systems and have already proven themselves among users, have been studied in detail. The main task addressed in this article was to find an answer to the question of the feasibility of using non-relational databases of the architecture of Cassandra, Redis, and Mongo depending on the characteristics of the system, or record information. Based on the analysis, options for using these databases for systems with a high number of requests to read or write information are proposed.