[sami] Scope [was Re: Welcome to SAMI and something you may like to know.]

[Melinda Shore <[e-mail]>]

I have to admit that I'm becoming more confused rather than less
as the discussions progress.

1) My understanding earlier was that we were talking about failing
    over services between existing VMs rather than copying a VM as
    part of a failover.  Is anybody actually doing that?  Aren't you
    concerned that you'll be copying the failed state?

2) If you're copying a live VM, along with its runtime  state
    (i.e. memory), things like IP addresses, MAC addresses, and
    so on will be moving with it.  If the VM stays on the same
    subnet presumably it will be behind the same middlebox.  The
    only thing I could think of that might become an issue is the
    MAC address being mapped to a particular switch port.  What
    network state do you think will need to move with a VM
    relocation on the same subnet?

It seems to me that there's some interesting work here regarding
transferring middlebox state when the service is being failed over,
not so much when a VM is just being shuttled around a subnet (and
I'm not sure that really happens, anyway).  That said, if what's
at issue is middlebox state moving that probably belongs in the
transport area rather than the ops area.  But first, let's figure
out what's actually under discussion.

Melinda