@Bing_Guo Did you try DADA2 on your dataset and find that it took a long time? If so we would like to know, as a dataset of that size (~10M reads spread across dozens(?) of samples) shouldn't take that long to process, even on a laptop.
It's hard to give accurate running time estimates from just a number of reads, as running time more closely tracks the number of unique sequences in your data (which in turn depends on how diverse the underlying community is, how high your chimera rate was, how high your error rate was...) but if you turn on mulithreading, I would expect that sort of dataset to complete within a day on a basic Amazon instance.
Again, let us know if that's not the case!