Re: CRUSH Algorithm

[Gregory Farnum <gfarnum@xxxxxxxxxx>]

The code is in ceph/src/crush of the gut repo, but it's pretty opaque. If you go to the Ceph site and look through the pages there's one about "publications" (or maybe just documentation? I think publications) that hosts a paper on how CRUSH works.

IIRC it's using the jenkins hash on the object name, but the part about selecting replicas based on a tree hierarchy is the interesting part. :)

-Greg 

On Monday, November 30, 2015, James Gallagher <james.np.gallagher@xxxxxxxxx> wrote:

Hi,

I was wondering what hash function the CRUSH algorithm used, is there any way that I can access the code for it? Or is it a commonly used one such as MD5 or SHA-1. Essentially, I'm just looking to read more information about it as I'm interested how this is used in order to look up objects independent from a list like other DFS

Thanks,

James

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com