부산대학교 도서관

Simulations of the desired system usually precede the development of P2P storage distributed systems. Justification for the development of the system is based on the results of simulations. Valid recognition of input parameters is essential for the quality output of such simulations. We collected data from the Faculty of information technology network, about node presence and patter of file generation. We discovered a small group of users with a similar pattern of behavior and file generation. Based on analyzed data, we developed a user model for this user group with parameters, such as the average size of user files stored on their storage locations and the pattern of file generation. After that, we compared the availability of a distributed storage system when using a coincidence matrix and a user clustering method, from our previous research. The aim of the research is to determine the moment of justification for using one of these two approaches in peer selection.