Most large-scale data server systems are having difficulties applying modern data usage patterns to such systems because recent data request patterns of users are sequential, and users tend to request up-to-date data. In this regard, customized systems are necessary for handling such requests efficiently. This paper deals with issues related to how conventional large-scale data server systems utilize memory, and how data are stored in storage devices. In addition, the paper analyzes data usage patterns of users, utilizing a cold storage system, and proposes a main memory system based on the analysis. This paper proposes a hybrid main memory system that utilizes DRAM and phase change memory (PCM). PCM is regarded as the next generation of non-volatile memory. Using a main memory that utilizes PCM, which operates similar to DRAM, and non-volatile storage, the proposed system improves the data processing efficiency. The paper also proposes an algorithm for processing data with the use of DRAM as a buffer. In addition, the paper proposes a system architecture with a tree-type block data and hash-type data block link. Moreover, this study compares the performance of an existing system with that of the proposed system using sequential and random data workloads. The results of the comparison show that performance improves by 10% when using a sequential data load, and remains almost at the same level when using a random data workload.
CITATION STYLE
Noh, T. H., & Kwon, S. J. (2019). Memory management strategy for PCM-based IoT cloud server. In Lecture Notes in Electrical Engineering (Vol. 513, pp. 69–77). Springer Verlag. https://doi.org/10.1007/978-981-13-1059-1_7
Mendeley helps you to discover research relevant for your work.