Zholtkevych G. Modeling data replication processes in distributed datastores.

Actual scientific-applied problem has been posed in the thesis – the problem of existence of general decision-making method, that would be the one of the base components on the stage of distributed datastore design, that allows to make larger size of distributed network and not to think permanently about possible clusters and nodes allocation, but to have specific criteria for building distributed datastore in the conditions of various system requirements and balance CAP-requirements depending on these needs. To solve the problem, existing methods of fulfilling CAP-requirements and distributed database replication processes are investigated. Models for distributed datastore that describe centralized and distributed data storing approach, distributed network were formed. This is the basis for the further research: it is developed the stochastic model to evaluate different components of CAP-requirements and metrics that impact these values. This gives the reason to form general recommendations for designing distributed datastore with balanced degrees of CAP-guarantees depending on the given business and system requirements; gives the ability to evolve methods of managing required degree consistency value (one of CAP-guarantees) on the needed level and gives possibility to monitor CAP-guarantees. Also, for that the method for balancing between consistent nodes is created and the algorithm is implemented