repository
대용량 저장장치에서 PCM을 활용한 효율적인 메모리 관리기법
- Alternative Title
- Efficient Memory Management Utilizing PCM from Large Scale Storage
- Author(s)
- Noh Taehoon
- Alternative Author(s)
- TaeHoon Noh
- Advisor
- Tae-Sung Chung
- Department
- 일반대학원 컴퓨터공학과
- Publisher
- The Graduate School, Ajou University
- Publication Year
- 2016-08
- Language
- eng
- Keyword
- PCM; Main Memory; Cold Storage; Data Pattern; Hybrid Archtecture
- Alternative Abstract
- Most of large-scale data server systems employ DRAM main memory and flash memory storage. However, it’s difficult to apply modern users’ data usage patterns to those systems. The reason is that recent user data request patterns are sequential and users tend to request up-to-date data. It means most users request data stored in servers in part. In this regard, customized systems are necessary for handling those requests efficiently. This paper deals with issues related to how conventional large-scale data server systems utilize memory and how data is stored in storage devices. Also, the paper analyzes user data usage patterns, utilizing 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 PCM. PCM is regarded as the next generation non-volatile memory. With the main memory that uses PCM which works like DRAM and non-volatile storage, the proposed system improves data process efficiency. The paper proposes an algorithm that process data with the use of DRAM as buffer. This system defines up-to-date data as hot data and uses it. This paper proposes a system architecture that has tree-type block data and hash-type data block link. And the paper compares performance of the existing system with that of the proposed system, using sequential data work load and random data workload. The result of the comparison shows that performance improves by 10% when using sequential data load, and remains almost the same when using random data workload. With the result, the system proposed in this paper proves more efficient than conventional systems that use user data usage patterns.
- Fulltext
- Appears in Collections:
- Graduate School of Ajou University > Department of Computer Engineering > 3. Theses(Master)
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
-
- License
- STATISTICS
- Total Visit :3,772,036
- Total Download :1,819
- Today View :8,537
