PageRank is a sparse mat and the eigenvalue can be obtained by multiply itself many times. But how to compute the general eigenvector in parallel? at least the eigenvector of symmetric matrix. It is very important when trying parallel PCA.