Distributed file systembased optimization algorithm

Database engines and file systems have been using prefetching and caching technologies for decades to enhance the performance of I/O-intensive applications. When future data access needs to be accelerated, prefetching methods often provide gains depending on the latency of the entire system by loading primary memory elements. Its execution time, where the data level prefetching rules are set, has to be much improved, as they are challenging to optimize, comprehend, and manage. This paper aims to introduce a novel distributed file system (DFS) model through dynamic prefetching, that includes four processes such as (1) Identification of popular files, (2) Estimation of support value for a file block, (3) Extraction of frequent block access patterns, and (4) Matching algorithm. At first, the input files are given to the first phase (i.e.), identification of popular sizes, where the popular files are identified. The support value of the file blocks that correspond to popular files is calculated in the second stage. Then, the extraction of frequent block access patterns is done in the third phase. At last, in the matching algorithm, the identification or prediction of frequent access pattern of the query is done by the optimized Neural Network (NN). Here, the weight of NN is optimally tuned by the Harmonic Mean based Grey Wolf Optimization (HMGWO) Algorithm.The proposed NN + HMGWO model produces reduced FPR values with good quality, which are 70.84%, 73.86%, 70.51%, 62.90%, 55.76%, 78.63%, and 73.86%, respectively, in comparison to other standard models like NN + WOA, NN + GWO, NN + PSO, NN + FF, FBAP, NN, and SVM. Lastly, the effectiveness of a chosen scheme is compared to other current methods in terms of delay analysis, latency analysis, hit ratio analysis, and correspondingly.