[clug] GFS - add some more abstraction [SEC=PERSONAL]

[steve jenkin]

Roppola, Antti wrote on 31/03/10 10:16 AM:
> Hi all,
> 
> So is it just me, or does GFS suffer from compulsive abstraction?
> 
> For a shared filesystem it seems to want layers for clustering, LVM and
> a good fistful of services too.
> 
> It seems a bit of overkill if one just wants a shared filesystem on a
> small private network. I find myself thinking fondly about NFS and
> wondering if GFS is really worth all that extra effort.
> 
> Has anyone else grappled with this?
> 
> Cheers,
> 
> Antti
> P.S. The GFS documentation also seems pretty patchy. There's a few
> behemoth documents that seem to assume you are building a data centre
> and not too much on why each layer is important.

Antti,

I was thinking exactly this the other day as I erased GFS from my
whiteboard checklist... (done/to do list)

The *one* useful thing I thought GFS does is allow every machine to
contribute spare disk space 'to the cloud'. But not automatically or
extensibly for what I saw.

I also makes a reasonable fist of allowing two machines to cluster their
disks and serve files to others using NFS... But its not the only solution.

If like ZFS, you could easily declare a disk slice to contribute to a
globally shared filesystem and then it all 'Just Works', that's the
ideal. Maybe I missed something in what I read/tried, or there's some
payfor version that does more like ZFS.

You gotta ask the question, "What would/does XXX do?" for XXX in {GOOG,
Amazon Compute Cloud, NetApp}

- GOOG because they might run the most servers or cluster nodes.
- Amazon EC^2 because it uses RedHat to do 'Cloud' stuff, and
- NetApp because they defined the NAS appliance market (& snapshots)

GOOG does 'whole system' redundancy and it's own extensible file-system
with multiple copies for redundancy.

AMZN - not sure just what they do/provide.

NetApp - started with FreeBSD around 1990. They cluster 'heads', do
active/active fail-over, share trays of disks and run their own WAFL
filesystem, normally configured as RAID 4 (four, sync'd writes of RAID
3), IIRC.


Hopefully those active users of Distributed/Shared Filesystems on the
list will correct my misperceptions...

cheers!
s