Basically, since you have a peptide the complexity of a system should allow to consider explicit approach (all atom waters, lipids etc.). I suppose we could discuss more details if you provide more info on aims of your research. The choice should be suited to your needs, as some times standard equilibrated MD is enough, in other cases it's worth to add some bias to sample rare events.

Nevertheless, one can't simply tell a priori how long simulations is needed to obtain relevant results - there's too much conditions. Most popular MD engines have pretty good support from community and can be found installed on many HPCs.

Hi, to stimulation of membrane you can use VMD software