On Thu, Dec 23, 2010 at 12:18 PM, Andrew Hume <[email protected]> wrote:
> we (some folks at work) have built two backblaze boxes
> (roughly speaking, a linux box with 45 2TB drives).

I'd be interested to know where you sourced the case from, or if you
built it yourself?  The cost for the backblaze pod is good enough
where I think we'd be willing to build one for demo purposes if we
could find reliable parts vendors.

> i think i know how to deploy such a beast, but wanted to check
> my understanding, which is that mdadm is the tool of choice,

There's DMRAID as well, though as far as I can tell it's still not as
fully baked in terms of support as mdadm and friends.  I assume that
since you're talking mdadm, using a cluster-like filesystem isn't
something you want to do?  There is also a FUSE port of ZFS as well as
a recently announced ZFS port that requires one to manually install it
due to licensing constraints.  I've never seriously looked at these
for deployment, so really don;t know if they're usable, but maybe
worth a mention.

> and that for performance and reliability, raid10 is the sweet spot
> (specificly, not RAID5).

Has worked for me though I've never benchmarked software raid5 vs
raid10 (normally that decision is made by availability constraints,
not performance, in cases where swraid is in use at all).

For mdadm raid10, read up on --layout options, as I'm given to
understand it can affect your performance as well as survival after
disk failures (eg, in the instance where a controller barfs and eats
the series of disks attached to it).

> does anyone have anything specific to say about mdadm,
> and the raid it produces, either good or bad?

As far as I can tell mdadm / linux raid seems more "twitchy" than the
standard run of the mill hardware RAID controllers and will tend to
kick a disk for any kind of slight issue.  That may also be due to my
use of it in places where there have been other hardware problems
related to the system, so YMMV.

Do you plan on having one fs per raidset or layering LVM on top of the
raidsets to generate randdomly sized fs'es?  The backblaze does
everything though their application via http, so I've always also
wondered how it would fare as a more traditional NAS/SAN provider.

Thanks!

-n
-- 
-------------------------------------------
nathan hruby <[email protected]>
metaphysically wrinkle-free
-------------------------------------------
_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to