[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ProgSoc] mac - network virtual-disk / sharing
On Friday 23 June 2006 4:31 pm, Nathan de Vries wrote:
] Scratch that. Apparently there's a kernel implementation of VBlade
] which will be much quicker for obvious reasons. I've been reading a
I have been reading up on this this afternoon, too -- thanks for
the pointer to it.
The VBlade stuff is just Coraid's commercial implementation of
a turnkey system. The AOE protocol is very much free software,
and implementations exist for Linux & Windows platforms, but not
Mac. Coraid's site suggests they'll have a client for Mac RSN, but
no timeframe specified.
I could presumably grab the source and compile it on a Mac, but
I've no idea how ugly it would be to get this up and running, or
whether I could use a 40gb file sitting on HFS+, rather than
re-partitioning their drives, etc.
iSCSI ... is hugely complex. As one of the plugs for AOE says, the
iSCSI spec is about 250 pages, and the AOE one is about 15.
But I'm not sure why they limited AOE to be non-routable -- it does
seem a severe limitation IMO.
A lot of the doco talks about how AOE saves on the *huge* overhead
of IP, even though in practical terms it's unlikely that this would
be the significant overhead they imply it to be. In reality you're
likely to be transferring fully populated packets, so the extra 30-40
bytes in a TCP header, compared to the 1500 bytes of data you're
shifting, isn't all that big a deal. At the very least it'd be nice
if the spec had allowed for the data to be routed, and had an
option to talk at a MAC (the other MAC) layer if you were feeling
particularly sensitive about performance on the day.
Jedd.
-
You are subscribed to the progsoc mailing list. To unsubscribe, send a
message containing "unsubscribe" to progsoc-request@xxxxxxxxxxxxxxxxxxx
If you are having trouble, ask owner-progsoc@xxxxxxxxxxxxxxxxxx for help.