Design of data backup on distributed memory system based on key-value store using hot/cold data management

Abstract

Recently, it has been observed that distributed memory systems are being preferred over traditional RDBMS system in several fields that require extensive data processing. However, it is difficult to store data permanently in a distributed memory system. In order to solve this problem, techniques such as snapshotting and logging are used. The disadvantages of using these techniques include high system load and increased file size in a large-scale distributed system. In this paper, we propose a scheme that applies snapshotting and data logging using hot/cold data management. It is possible to reduce system load and process large-scale transactions using this scheme.