Research On Ontology Based Storage System Management

After several decades'evolution, storage systems have grown from early simple peripheral equipments to complex installations that contain even thousands more computers. Related research is also promoted to a comprehensive area that involes architecture, performance, security and much more. Modern storage systems reveal their complex nature in many ways including the design, implementation and more specifically, the application environment. Furthermore, these growing complexities impose an ever increasing pressure to system performance, dependability and security. Although storage management practitioners progress much in directions including standardization and technology, the management of complex storage installation still depends largely on the knowledge and experience of seasoned administrators. Problems remain big in both management cost and efficiency.For facing this challenge, this research brings recent progressing work from Semantic Web and Knowledge Representation to storage management. The work start with a survey on ontology languages and ontology construction methods, the essential part of ontology engineering is reusing existing ontologies as far as possible. Then, based on the requirement of storage management, a reuse plan is proposed and applied to the design of storage management ontology (SMO), for covering general concepts in storage systems such as disk array and file systemThere are two problems in the creation of SMO domain ontology:expressivity and tractability. Expressivity determines whether the ontology is able to depict real-world storage management scenarios, yet over-detailed concept model will affect the inference efficiency. Thus a domain-restricted model conversion is proposed, which limit the model reusing to an area defined by the targeting storage systems. The industry-approved and well-practiced storage management specification ensures the completeness of resulting ontology, and the description logic based model translation guarantees the consistency. After that, SMO domain ontology is used in a series of tentative problem analysis to verify its practicability.Before SMO can be utilized in storage management, connection should be founded between general knowledge and particular management operations. This involves a framework that composed of ontology repository and inference engine, and task ontologies for describing specific management process. First, by adopting and integrating matured technologies from ontology research area, the framework is designed and implemented, and then S MO can be used to interact with actual management environment. Second, SMO task ontologies are created aimed at disk array management and file system management. According to the evolution and expanding application of disk array technology, the related concept in SMO can therefore be extended, and introducing the management knowledge to expanded application scenerios effectively. Considering the application locality of file systems, an application-oriented management method AO2 is proposed. AO2 demands several key information to be analyzed together, therefore cannot be realized by simply running traditional management tools. Dedicated task ontology is then constructed and used in problem inference to fulfill this mission.There is still one key step towards applying SMO to practical storage management. This is the execution of management operation. Besides that, the outcome of such management methods should be evaluated by experiments accordingly. So 3 different experiment cases are deployed on several typical operating systems and file systems. They are application-oriented defragmentation, namespace and prefetch, each case is equipped with required toolset for carring out the management operations. The results show that, compared with traditional management tools and methods, the SMO-based management is more agile, autonomous and effective.This research unleashes the potential of using knowledge engineering technologies in storage management, provides a method for dealing with the increasing storage complexity. On the other hand, it brings semantic web and knowledge representation technologies into a new application area. With the development in both areas, a deeper integration of storage and computing intelligence is definitely predictable, leading to a promising future of better information storage, access and management.