Re: [Aoetools-discuss] ATAoE client/target compatibility - a nightmare

[Sam H. <sa...@co...>]

Hi Mark,

Questions about coraid are best suited to su...@co... as this is
simply a community forum where some of us coraiders hang out.  I'll
respond here, but let's be sure to keep the aoetools list from becoming
a place to discuss coraid.

> What's the state of ATAoE? I read the spec and well - it is simple.  
> But does it sufficiently describe all that's needed to get different  
> ATAoE targets / clients to work seamlessly together?

As far as we know, it does, from a developer's point of view.  As a user
you'd likely be best served reading documentation on how to use the
tools built to utilize the protocol than you would be analyzing the protocol
itself.  That being said, maybe you'll want to look at packet traces to, eg,
track i/o patterns.  In that case knowing what goes into the packets
can be helpful.

> Are there any known compatibility issues on the AoE landscape: will a  
> Coraid appliance behave/respond exactly the same to clients than a  
> vblade target?

There are performance differences, obviously, but as far as the aoe
protocol goes there should be no difference.  Vblade was written to
serve as an example aoe target implementation for developers wanting
to make targets and as a freely available aoe target for users to
become comfortable with aoe.  Many of our customers started with
vblade installations, outgrew them, and came to us for faster/bigger
targets.

> What is Coraid's commitment on making changes/enhancements to the  
> ATAoE spec public?

We haven't made any fundamental changes to the protocol in quite a
while.  The last minor change was updating the config-query response
so the target could report its max sector request count for using
jumbo frames.  When we do find it necessary to make a change, though,
we immediately update the protocol document.  We respect the openness
needed to participate in the open source community.

We've occasionally come up with aoe vendor extensions that aren't in
the aoe protocol document because they don't belong there.  One is the
mac mask command (documented in aoemask from the SR support page).
Anyone can define vendor extensions, though; that's why we reserved
the command space for them.

I hope that helps!

Cheers,

Sam