Desain dan Implementasi Sistem Terdistribusi dengan Model Replikasi Multi-master pada Basis Data untuk Optimasi Keandalan Sistem
Laday, Ravie Kurnia
MetadataShow full item record
An important factor in the usability of an information system service is security and reliability guarantee. Computer network reliability is defined as the ability to optimize functions in a particular period. Performance, availability, and security are factors influencing reliability. In term of ensuring the reliability, replication technique is one of the mechanisms that copy and distribute data from one database into another. It synchronizes between one or more databases that guarantee data consistency. Multi-master replication is a process that allows any computer can write and read data in a master database. When there is a change in one database master then other master will be updated automatically, and vice versa. A distributed system is a form of the system architecture in which computers are autonomous can communicate and share resources regardless of where the computer was located, and the platform used. Problems in both data processing in data loss and the backup method are still manual so that existing systems less reliable. Seeing these problems need to be constructed a prototype system that is more reliable both in thems of availability or system performance. The study was conducted at the Insistute of Science and Technology Al-Kamal using sample data and existing architecture. The existing system was engineered in a prototype that utilizes a distributed load balancing technology for the system. Furthermore, the research built a backup database with multi-master replication model. Multi-master replication method can improve the availability of the system model. The research resulted that data changes were not lost as long as the system could guarantee their connections. Performance distributed systems using a load balancer can maintain server availability. The system takes 0.731-second and 63.56 to show login page and student data. The number of connections did not affect tes-duration, however, it affected the reply time. F-value 0.11 form test duration and 0.07 for reply time proved the conclusion. The result of this research and model will be proposed as a new model to replace the existing system model.