The analysis of a distributed loop switching system with multirank buffers is based on the Markov process. The mergeability of the process is proved for a system of single-rank buffers, and generalized to systems of multirank buffers. The results for single-rank buffers are found to be in agreement with already established results. The solution for double-rank buffers, the only multirank buffer system studied in detail, is compared to the results obtained from simulation. A complete agreement is found within 5 percent, over the full dynamic range. General and normalized curves are obtained that enable us to predict the performances of double-rank buffer systems.