[Sis-ams] spec changes
Scott Burleigh
Scott.Burleigh at jpl.nasa.gov
Fri Sep 30 16:07:32 EDT 2005
Hi. I haven't had time to post to this list a detailed recap of the
discussions in Atlanta early this month, and I can't go into a whole lot
of detail right now either -- that will show up mainly in the next
edition of the Concept Paper, which I'm working on now, and in
discussion on this list after I get that draft posted. But I want to
talk a little about the two main types of changes to AMS that came out
of the meetings, because there are broader implications to one of them.
First, in talking with Roger Thompson about AMS features that MOIMS/SMC
will need, I came to the realization that we can easily make the AMS
"zone" and "role" notions powerful enough to minimize the need for
elaborate subclassing or filtering of message subjects. Specifically, I
think Roger wanted the ability to subscribe either to (here I'm making
up a dubious and fictitious example) "instrument state" messages
published by the UV spectrometer on Deep Space 5, or the "instrument
state" messages published by all of the instruments on Deep Space 5, or
all of the "instrument state" messages published by the UV spectrometers
on all flying spacecraft, or all of the "instrument state" messages
published by all instruments on all flying spacecraft. You could do
this with message subject naming, e.g., "DS5.UVS.instrumentState", but
that makes subscribing to all "instrumentState" messages tedious.
Alternatively, though, you could have a hierarchy of zones (e.g., the
"all flying spacecraft" zone could contain zones named "DS5", "DS6",
etc.), and when subscribing to messages you could specify not only
message subject "instrumentState" but also the zone ("DS5") and
functional role ("UVS") of the AMS nodes whose published messages you
want to receive. You could similarly constrain message invitations,
with the effect that you would accept messages only from nodes
performing specified application roles and registered within specified
zones. And, for that matter, we could add a new "announce" message
transmission model in addition to "publish" and "send": when you
announce a message you are telling AMS to send it to all nodes
performing specified application roles and registered within specified
zones, rather than to a specific node or to all subscribers to a
specific subject. I think this turns out to be a surprisingly small
extension to the protocol and the implementation that offers a lot of
expressive power.
The other major change came from sitting in on the SOIS/TCONS meeting
where we discussed the TCONS QOS model. It occurred to me that AMS
could utilize both the TCONS resource reservation "channel" notion, and
also other underlying transport service QOS mechanisms like diffserv, by
adding a "flow label" to subscriptions and invitations and letting
transport service adapters map flow labels to channels (etc.) according
to preconfigured, managed rules. This seems like a really easy and
general mechanism for meeting some of the AMS QOS requirements.
Along those lines, though: given this additional QOS dimension, do we
still need delivery deadlines as part of the AMS QOS design? The reason
I'm having second thoughts is that MTS really hasn't been absorbed into
AMS: it is turning into a separate, alternative API to AMS
functionality, complementing the native AMS API, in much the same way
that RPCs are an alternative to sockets as an API for accessing TCP/IP
functionality. MTS is already doing things like managing connections,
which are not an AMS concept; maybe the delivery deadlines and related
features identified in the MTS spec similarly ought to be implemented in
MTS rather than in AMS. This would have the advantage of somewhat
simplifying and streamlining AMS, making additional implementations and
testing somewhat easier. I haven't heard a lot of demand for these
features from the C3I or SMC communities; maybe they are sufficiently
SOIS-specific to make them more appropriate for MTS than for the
underlying AMS. Anyone have any opinions on this?
One last note: for those who are interested, I have posted to the AMS
CWE site the AMS status slides I was talking from at the Atlanta
meetings. Let me know if you've got questions on 'em.
Scott
More information about the Sis-ams
mailing list