Parallelizable numerical methods for solving large scale DAE systems are developed at the level of differential, nonlinear and linear equations. For this the subsystem-wise structure of the DAE systems based on unit-oriented modelling is explored. Partitionings are used to parallelize waveform relaxation and structured Newton methods. Initial values are computed with a modified Newton method. To solve large sparse systems of linear equations a special Gaussian elimination method is used. The algorithms were implemented on a CRAY C90 vector computer, as well as on both, moderately parallel CRAY J90 vector computers and massively parallel CRAY T3D machines. The methods were tested using several real life examples. (orig.)Available from TIB Ha...