Traditional studies on disk arrays have focused on parallelism or load balance of disks. However, this paper reveals that the independency of disks is more important than parallelism for concurrent reads of large numbers of processes in striped disk arrays, whereas parallelism is significant only for concurrent reads of small numbers of processes. We investigate and evaluate how much aligning sequential prefetch in strip or stripe boundaries affects the I/O performance by comparing with the proposed two types of sequential prefetching schemes, both of which are implemented in Linux kernel 2.6.18. In the experiments in this study, the combination of our schemes outperforms the original sequential prefetching of Linux by 3.2 times for 128 clients and 2.4 times for a single sequential read.