Supernova 2.0, printed on 12/06/2019
Supernova pipelines run on individual Linux systems that meet these minimum requirements:
Supernova is thoroughly tested at default settings. You can adjust the number of CPUs used, and it may be beneficial to do so, but some subsystems will not take advantage of more than 32. We also strongly discourage attempting to use fewer than 16.
We recommend that servers have hyperthreading turned off, however we only make
this recommendation because that is our internal configuration, and it works well for us. We have not tested the alternative.
256 GB RAM for a human at 56x coverage (tested successfully on eight datasets; probably between 1 and 10% of human datasets at 56x coverage require 10-20% more memory)
256 GB RAM for most 56x datasets for genomes ≤ 1 GB that are only modestly repetitive, however it is not possible to make guarantees; note that many small genomes would likely run on servers having only 128 GB
512 GB RAM for some nonhuman datasets
> 512 GB RAM for some nonhuman datasets; we have no examples but expect that some genomes larger than 3.2 Gb could require this
Please see observed memory requirements for test assemblies.
No other large processes running on the system, except when all processes have their memory capped appropriately. Supernova has an option (described in Assembly Process) for capping its memory usage. Running multiple large user processes on the same server will have an unpredictable effect on run time.
In addition, Supernova must be run on a system with the following software pre-installed:
bcl2fastq 2.17 or higher is preferred and supports most sequencers running RTA version 1.18.54 or higher. If you are using a NovaSeq, please use version 2.20 or higher. If your sequencer is running an older version of RTA, then bcl2fastq 1.8.4 is required by Illumina.