Design And Implementation Of Object-Based Storage File System

As a crucial component of Object-Based Storage System, Object-Based storage file system has become research focus of distributed file system. By virtue of its advantage in scalability, security and performance, Object-Based storage file system made itself first choice in high performance computation, life science, energy and other fields.Design and implement an Object-Based storage file system named HUSTOBS used in Object-Based storage system. HUSTOBS file system consists of Object-Based storage device, metadata server and client, just like most popular distributed file system. The three components are interconnected with Gigabit Ethernet, and work together to complete a data operation. Unlike traditional storage interface, HOSTOBS uses "object" as its basic transmissions unit. Object is of variable length which inherits block interface's high performance and file interface's cross-platform data sharing. All objects has attributes which reflects some characters of the object. "Object" creates much improvement space in scalability, security, management costs and performance. HUSTOBS is fully compatible with T10 SCSI OSD standard: Device is based dedicated embedded system; Client is implemented in Windows and Linux respectively, and can provide API access in both environments, furthermore, both clients can use device as local disk (virtual logic volume under Windows and virtual directory under Linux). This way it facilitates different demands.Object-Based storage file system is optimized in data transfer. HUSTOBS realized aggregated write and prefetch in client, which makes good performance test result. HUSTOBS tracks clients' access preference in a short time span by defining a "prefetch" attribute page for every object and realize a self-adaptive dynamic prefetch policy, this way prefetch hit rate and the whole performance is improved.