SEQanswers

SEQanswers (http://seqanswers.com/forums/index.php)
-   Metagenomics (http://seqanswers.com/forums/forumdisplay.php?f=29)
-   -   Using MetaVelvet, what hardware requirements? (http://seqanswers.com/forums/showthread.php?t=34604)

rjoyons 10-18-2013 02:12 AM

Using MetaVelvet, what hardware requirements?
 
Hi everyone,

I just received my very first set of shotgun-metagenomic sequencing results. They are illumina pair ended reads produced using the 2x250bp method. I have 26.5 million reads.

My question is about the memory requirements of metavelvet.. In the manual it says the program requires "At least 12Gb of RAM (more is no luxury)" But they go onto use a computer with 48Gb of RAM to produce their example data in table 1 and the "required memory" for each assembly in tabel 1 is up to 72Gb.

I'm finding this quite confusing! What happens when the memory requirement exceeds the amount of RAM on the workstation? Will the program crash or start making temporary files on the hard drive? Can you really use a computer with only 12Gb of RAM?

Any advice would be greatly appreciated!! Please help!

student-t 06-01-2015 07:26 PM

The approximate memory requirement is: number of nodes * 416 bytes because 416 bytes is the memory required for a node data-structure in C.

bastianwur 06-04-2015 07:10 AM

Quote:

Originally Posted by rjoyons (Post 119173)
HWhat happens when the memory requirement exceeds the amount of RAM on the workstation? Will the program crash or start making temporary files on the hard drive?

Default behaviour: First it'll use the swap, and if the swap gets nearly full, then the kernel will kill the program.

Do you really have only 12 GB available?

Apparently the Megahit assembler is really memory efficient, you might want to try that one.

student-t 06-04-2015 04:48 PM

This is also known as virtual memory, where the hard-disk space is being used for paging.

Brian Bushnell 06-04-2015 05:04 PM

I'm hoping this project has been assembled by now. That said, Megahit is indeed quite efficient and we are now using it for metagenomes, with better results than Soap using a small fraction of the time and memory.

If you run out of memory with assemblies, particularly metagenomes, it can be helpful to quality-trim, error-correct, and normalize the data first. This can greatly decrease the resource requirements (both time and memory) by reducing the volume of input reads and unique input kmers.

student-t 06-04-2015 10:39 PM

Brian, do know any paper that reviews it? I'm interested in knowing more.


All times are GMT -8. The time now is 11:32 AM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2018, vBulletin Solutions, Inc.