Thanks Dr. Paul R. Yarnold for your concern . now I want to explain more about my question : I working now on stochastic volitility model and I have to estimate all parameters in this model . To do this,first  I did discritization of the model to convert it from continuous to discrete . Second : to get the estimations of the parameters I used maximum likeihood fumction but the covariance matrix is not easy to compute so as altrenative I use innovation algorithm .

my problem : many of my calculations depend on fractional Gaussian noise process and I have to apply my program on size of 100,200,300,400 and 500 for 100 replications. This take very very long time. For that reason I thinking in dividing the fractional Gaussian noise into equal subprocesses. Then apply my program on each subprocess to get the estimations of the parameters. Finally, I will compute the average of the estimations of each parameter. This idea take too shorter time .