using velvet for big databases

[sheerlik]

Hello,
I'm trying to assemble large amount of reads (1.5 T of reads) using velvet 1.2.08 (and using metavelvet after using velveth and velvetg). The process always terminates at velveth. I would greatly appreciate if you could shed some light on the reason for these terminations. At the log file (in file /home/a-m/itzhakm/Runall_samples_51.o348823) there is an error message :
> "velveth: Can't calloc 18446744072189905506 char *s totalling 18446744061552382736 bytes: Cannot allocate memory[meta-velvetg] Error: There is no 'Roadmaps' file in the directory: /home/a-m/itzhakm/Assembly/all_51_dir You should run 'velveth' before running 'meta-velvetg'"

Thank you,
Sheerli

[dpryan]

It's trying to reserve 18 exabytes of memory. That's going to be a problem.

[sheerlik]

I know. i don't want to decrease the k-mer (51), I'm trying to use khmer for filtering. Do you know this software?
Thank you,
Sheerli

[GenoMax]

Unless you try the assemblies with subsets of data this will always be a problem. What sort of genome are you working with that you needed 1.5 T (is that terabases or trillion) reads?