RE: mass storage recommendations

On Wed, 2006-05-31 at 10:07 -0500, Richard Pickett wrote:
> > when u have 100TB from 750GB disks, raid5 is not enough for you. u
> need
> > to look at raid6.
> 
> Ok, thanks!

there is a technical paper from Netapp about the probability of raid5
and raid6. it is quite interesting read.

basically when u have so many disks and each are so large, u should look
at raid6. especially u said it is mainly for secondary and not have
critical performance need on it.

> 
> > > I was thinking that, only with the 8 (or maybe 10) disks per
> physical
> > > box, then tying them together with one box in the front, and it
> doing
> > > raid5 on the enbd disks it pulls in. In this way it would actually
> take
> > > 4 disks (two on two different systems) or 2 whole systems to fail
> before
> > > the data wouldn't be accessible (or the front end box, which that's
> > > simple enough to fix, we just drop another front end in place enbd
> the
> > > backends into the new frontend).
> > 
> > raid5 over raid5 will be quite slow.
> 
> That was a concern of ours. Especially when we're talking about so many
> enbd disks, we were even thinking about segmenting several machines on
> separate gig switches to help reduce the network bottleneck.

yes, this must be a issue to consider.

> 
> > will u front end become a performance bottleneck? what kind of io
> > workload u are expecting on this front end?
> 
> Mostly it's just read access. It's secondary storage and is only
> accessed when the primary sources (which are remote and scattered across
> the world) are either inaccessible or destroyed.

if this is the case, i think make each storage node a ENBD or ISCSI
nodes, then run file system like lustre or ocfs or gfs in front of it.
maybe easier.



> 
> Thanks for your reply.
> 
> Richard W. Pickett, Jr.
> President, CSR Technologies .com, Inc.
> [email protected]
> Office - (270) 746-0324
> Cell    - (270) 303-9154
> 
> 
> 
>