Specs on what this should take to run?

[Joseph Chase]

Hello, I am thinking of implementing this on a server and am wondering if there is a location for requirements needed to run this effectively (will have around 10TB of phewas data).

-Joe

[Peter VandeHaar]

How many variants?  How many phenotypes?

I recommend loading on a big machine (50+ cores) or a cluster for that dataset size.  The only step that uses much RAM is the qq plots, so make sure that each thread has num_variants*16 bytes of RAM during that step (by setting num_procs). 

To start, you can try with just the first 100k variants, to get an estimate of how long it’ll take. 

Serving a pheweb is fine on 1 core + 1GB RAM, but 8 cores will serve faster.

On Fri, Jul 22, 2022 at 1:54 PM Joseph Chase <jcon...@gmail.com> wrote:

Hello, I am thinking of implementing this on a server and am wondering if there is a location for requirements needed to run this effectively (will have around 10TB of phewas data).

-Joe

--
You received this message because you are subscribed to the Google Groups "PheWeb-UMich" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pheweb-umich...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pheweb-umich/5dc07efe-f6db-49a8-9026-254afa4605bcn%40googlegroups.com.