Research And Implementation Of Metadata Server Cluster Based On LDAP

With the explosive growth of digital information, storage systems are getting larger and more complicated in order to store increasingly diverse types of data. Object-based storage technology, which can provide high quality and scalability for storage system, is becoming the foundation of large-scale distributed storage system and a new research hotspot of storage field. Acting as a core of Object-based storage system, metadata server subsystem is the key to the data distribution. However, with respect to its' service efficiency, availability and scalability, traditional single metadata server cannot have efficient management of TB or even PB amount of data in large object storage system. As a result, studying how to construct metadata server cluster, so as to provide a sound metadata service for object storage system, is an important and urgent problem.On the basis of research on the architecture of Object-based storage, and in view of Lightweight Directory Access Protocol's good behaviour of distributed directory service, a metadata server cluster based on LDAP has been presented. The feature of this cluster is the separation of service support and request processing. There is no centralized point and provide completely distributed management. In order to implement this distributed management, it uses a metadata partition strategy based on hashing, which takes object's ID as the hash key. Besides, a strategy of dynamic feedback on load balance is designed, which makes the LMDS Cluster take use of the resources in a proper way. To meet the requirement of the consistency of data copy, a two-phase synchronization protocol is proposed. And this process of synchronization is melt into the process of the metadata requests to guarantee the real-time online synchronization.At present, the prototype architecture of LMDS Cluster has been implemented. Its testing results show that this architecture can respond to the concurrent requests from the clients with a stable and good performance, which embodies the capability of metadata management aimed at large-scale storage systems.