Experiences with Novosort

[zee]

Hi, We recently had a user send us information on their experience with novosort on large BAM files.

Picard was previously used for sorting, but (due to the underlying Java virtual machine?) without multiple threads, he switched to novosort. All runs were done on a 24 CPU machine.
Below are the running times for the same 10 samples (~120M paired-end reads for each sample, origin is human):

Picard SortSam:
Elapsed time: 47.47 minutes.
Elapsed time: 38.48 minutes.
Elapsed time: 44.57 minutes.
Elapsed time: 42.29 minutes.
Elapsed time: 43.65 minutes.
Elapsed time: 39.21 minutes.
Elapsed time: 39.05 minutes.
Elapsed time: 37.75 minutes.
Elapsed time: 39.30 minutes.
Elapsed time: 41.41 minutes.

Novosort:
# Elapsed Time: 4.1 (min.)
# Elapsed Time: 3.2 (min.)
# Elapsed Time: 3.9 (min.)
# Elapsed Time: 3.0 (min.)
# Elapsed Time: 3.7 (min.)
# Elapsed Time: 3.0 (min.)
# Elapsed Time: 2.9 (min.)
# Elapsed Time: 3.4 (min.)
# Elapsed Time: 2.8 (min.)
# Elapsed Time: 3.2 (min.)


Would anybody be able to share information for a similar run with any other parallel SAM sorting program? We would love to get more feedback from the community.
Novosort is free to use and can be downloaded at www.novocraft.com .