Research On The Key Technologies Of Storage Systems Based On Redundant Intelligent Storage Channel

Traditional host-based storage framework tends to be networked gradually and the separation of computing and storage is more and more obvious. In distributed storage environment, the tolerance ability and performance of networked storage systems becomes one of research hotspots. The concept to devise about Redundant Intelligent Storage Channel (RISC) is to intensify the research of key technologies of tolerant storage systems.To improve the performance of IP storage system, the concept of Redundant Intelligent Storage Channel is proposed. The channel technology realizes the multiple network-card aggregation, intelligent fault-detection, and virtual I/O interface to reduce I/O-processing time and improve the availability of the storage system. The availability model of RISC channel to provide networked storage service is studied and is used to analyze the relevance of availability of RISC-channel service with the number of channels, channel-beared payload, and buffer size; the results of the simulation of the models and tests show that these parameters have the different roles on service availability.The distributed tolerant storage system is constructed by RISC channel technology. The research on this system discusses the Lease protocol model of buffer consistency in accordance with requirements of caching metadata information; from the features of the fork-join model associated with Lee model of disk arrays, the queue models of tolerant storage exemplified by iSCSI RAIDO-land RAID5 are developed; the system-responsed time is computed in accordance with the concrete parameters of the model to obtain the performance curves to show that to control stripe size will reduce the average system-responsed time. Iometer tool is used in Linux development environment to obtain the results of the performance tests of the iSCSI-based tolerant storage system.Based on the six properties of ideal data layout, single-point error-correction, parity distribution, reconstruction distribution, mapping with high efficiency, large-size write optimization, and the maximum of parallel, pointed to by Holland&Gibson, the data layout in RISC-based tolerant stoarage systems on influence of reliability and storage-accessing efficiency is studied by examples about four types of mirroring:basic mirroring, Group-rotate Declustering, Interleaved Declustering, and Chained Declustering as well as iSCSI Mirroring and iSCSI RAID5. Based on the relationship between data layout and reliability, several performance-optimizing fault-tolerant layouts are designed. Considering network failures, the Markov relaiblity mode of the performance-optimizing fault-tolerant layout against double faults is desined and is easy to compute its reliability.A method is proposed to analyze the time-varing property of the reliability of RISC-based tolerant storage system by using the theory of fuzzy mathematics. The method develops the Markov inspiration models of iSCSI RAID5 and RAID5x by using the triangle fuzzy number in order to obtain the time-dependent features of their tolerant stoarage systems. Meantime, the hypothesis of fuzzy status and probability is proposed and the state-transition for the tolerant stoage systems is analyzed by examples of four mirroring-like organizations. The simulation results obtained by time sampling evaluation show the reliability difference of storage systems of such four mirroring-lik organization.The research on the above key technologies is able both to improve the tolerant ability of storage systems and to keep the much more performance superiority. It has the positive value on the massive storage application.