A design study for network based storage systems and performance evaluation
Hui Guo, Jinli Zhou, Lihui Yang, Shengsheng Yu
Abstract
Hui Guo, Jinli Zhou, Lihui Yang, Shengsheng Yu
Abstract
It is well know that network-attached storage (NAS) systems have several advantages compared with traditional storage systems. NAS offers new possibilities in building a distributed storage system, providing direct data transfer between the clients and storage disks. We discuss the underlying technology trends that are leading to high performance network-based storage, and provide a description of the design of our NAS system, HUSTserver (Huazhong University of Science and Technology server), from the software and hardware architecture point of view. RAID (redundant arrays of inexpensive disks) is a key technology leading to high performance and reliability in the I/O subsystem. We introduce a software RAID prototype, RAIDframe, as our system disk volume manager. In addition, we report on the performance obtained at each layer of the system, including the RAID I/O subsystem and the overall system capacity.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
It is well know that network-attached storage (NAS) systems have several advantages compared with traditional storage systems. NAS offers new possibilities in building a distributed storage system, providing direct data transfer between the clients and storage disks. We discuss the underlying technology trends that are leading to high performance network-based storage, and provide a description of the design of our NAS system, HUSTserver (Huazhong University of Science and Technology server), from the software and hardware architecture point of view. RAID (redundant arrays of inexpensive disks) is a key technology leading to high performance and reliability in the I/O subsystem. We introduce a software RAID prototype, RAIDframe, as our system disk volume manager. In addition, we report on the performance obtained at each layer of the system, including the RAID I/O subsystem and the overall system capacity.
Key concepts: RAID, Computer science, Storage area network, Converged storage, Operating system, Reliability (semiconductor), Computer data storage, Embedded system