Data mining prefetching algorithm for multilevel storage system for the classification of high resolution data

Improving the performance, cost and energy efficiency of a hybrid storage system is too much of a challenge, because performance, cost and energy are often conflicting goal. As a huge hybrid data center, use of the online satellite image storage system EROS (Earth Resources Observation and Science) is an example of a model data centre that is not viable. Images are converted into a data unit and stored in a large scale within a large server. To assemble fast and process enormous data, the pre-fetching algorithm DMPAS was proposed to optimize a hybrid storage system. A hybrid storage system usually has three types of data storage components, Tape Drives (TD), Hard Disk Drives (HDD) and Solid State Drives (SSD). According to the performance analysis, TD has a slower data transfer rate than HDD and SSD. Though, SSD is faster than HDD and TD, it is quite costly in the market. The main idea for implementing the DM-PAS algorithm is to transfer the most used data to the Solid State Drive and to use archive data to the Tape Drive. This technological improvement will helps the data center increase performance and minimize the searching and seeking time. The algorithm was tested in a real world computer system. It is quite impossible to test the algorithm in the real world; but, to prove the algorithm true, a simulator had to be implemented and every possible variable had to be changed to get the better result