Data distribution is one of the most essential architectures of any serving network. Data storage and its retrieval depend a lot on how the data is organized in the distributed environment. With the fast development of technology, the requirements of users have also changed. A user who was stationary earlier has become mobile now and requires access to the data from anywhere in the world. An unorganized data structure will result in output delay in the network and may further result in user migration from one service provider to another service provider. Data fragmentation is one of the most essential parts when it comes to data storage. Organized data always gives convenience to others to use it conveniently. Due to the vast collection of data extraction of information in a fast manner is very complicated. So, to achieve performance in a distributed system an optimal strategy is required to overcome previous lapses and serves the maximum number of users in a wide geographical network. This research paper proposes a novel relative based fragmentation method that analyses the attributes of the data in relative architecture and is helpful to achieve query performance with better speed and accuracy. To assess the current proposed work a comparison has been drawn between k-means dependent cosine similarity measurement and hybridization of cosine and soft-cosine partition methods for data partitioning. Mentioned results in the article shows that the proposed similarity-based threshold segmentation method outperforms the existing in terms of partitioning strategy, precision, and recall parameters to achieve performance.