[OpenAFS] OpenAFS speed - some benchmarks

[Derek Atkins]

Rodney M Dyer <rmdyer@uncc.edu> writes:

> At 03:25 PM 6/25/2003 -0400, Ken Hornstein wrote:
> >A uid namespace?  I don't understand.  From what I've seen, there is only
> >a "userid" namespace.
> 
> I'm confused.  Kerberos is for "authentication" not "authorization".
> What I'm asking is what is the "authorization" method used under
> NFSv4?  Normally a NFS server maps "uids" of "authenticated users" to
> files right?  Well, for multiple NFS servers using Kerberos
> "authentication" there needs to be some way of mapping "authenticated"
> users to "uids" on those systems...right?  Since all the NFS servers,
> even ones run by individuals, fall under the same Kerberos
> authentication namespace, then they all need to use the same "uid"
> space too...right?  Otherwise, how do you perform a chown?

No, they do not... The NFSv4 RPCs could perform a translation for you,
theoretically.  AFS certainly doesn't need to keep UID space the
same across multiple AFS Cells.

Granted, it _looks_ better if you keep the uid space the same.  But
theoretically if you have a centralized Kerberos infrastructure you
probably also have a centralized Hesiod/LDAP/NIS which implies a
centralized uid namespace.....

> Rodney

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available