I have a quite large data set and will get a even larger one. I wish the analyzing process could be more efficiently. When i run mothur on a cluster with 96 CPUs (8 cpus on each node), it only uses 1 CPU and take quite long time. Is it possible to use more CPU? And how to set it up?
Yes it is possible, although you might only be able to access 8 cpus at a time with the current version. The align.seqs, chimera.seqs, and dist.seqs all have a “processors” option, which will allow you to use multiple processors. See those wiki pages to see how it’s done. We are working on parallelizing other commands. Also, we are working on changing how we parallelize things so that we use mpi so that you could then use all 96 cpus.
Hope this helps,
Is it really limited to 8 processors? I’ve been scheduling mother on a 32 processor system and haven’t seen any error message…
sorry, i think he might be limited to 8 since i don’t know if it can communicate across nodes the way it is currently programmed. are you sure that you are really accessing all 96 cpus?
I see. However, in my computational process, the speed limited step is the OTU classification. Is it poissible for this process to use multiple CPU under the commond model in the furture version?
yeah, we’ve got plans for this, but it may be a few months off. i’m not aware of many (any?) parallelized algorithms for heirarchical clustering algorithms.
I haven’t really checked to see how many processors it was using (I request 32 cores total, so that’s 8 quad-core processors).
The 1.9 release includes a mpi-enabled version.