I tried to multiply the large matrices (million by million). But, Hbase daemons are crashed during random matrix generation. So, I think it also should be parallelized. -- Best Regards, Edward J. Yoon @ NHN, corp. edwardyoon@apache.org http://blog.udanax.org