Running AUGUSTUS on a single computer is relatively easy. Nevertheless, depending on the number and length of sequences and the parameter values, gene prediction can be computationally very intensive. For example, the application to the human genome with its 3 billion base pairs can take up to several weeks. A possibility to speed up the computation without modifying the program source code is to parallelize the computation. This can be done by splitting the input file into smaller files which contain fewer sequences, and to run the program on these input files separately. Long sequences can be split into several smaller sequences if a suitable overlap according to maximum gene size is used. Data parallelization provides the possibility of distributing the computation across several machines, for example the nodes of a local cluster. Typically, these nodes have a shared file system, so the handling of input and output data is easy. However, additional effort is necessary for handling the batch system, and for checking if all submitted jobs are executed properly.