[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [cosmos-dev] RE: COSMOS DC: Review of Indexing examples
|
I am writing a wiki to capture the conclusion
from yesterday's meeitng: http://wiki.eclipse.org/COSMOSBrokerSpecification
I haven't finished writing. So
refresh the page from time to time to get my lastest updates.
_________________________
Hubert Leung
IBM Toronto Lab
hkyleung@xxxxxxxxxx
905-413-3382
Mark D Weitzel <weitzelm@xxxxxxxxxx>
Sent by: cosmos-dev-bounces@xxxxxxxxxxx
08/21/2007 09:27 AM
Please respond to
Cosmos Dev <cosmos-dev@xxxxxxxxxxx> |
|
To
| "Mohsin, Jimmy" <Jimmy.Mohsin@xxxxxx>
|
cc
| "Simmonds, Martin" <Martin.Simmonds@xxxxxx>,
Cosmos Dev <cosmos-dev@xxxxxxxxxxx>
|
Subject
| [cosmos-dev] RE: COSMOS DC: Review of
Indexing examples |
|
OK....
So we seem to be getting some closure on a few things. Here is what
I think are the next set of things for us to do...
1. Valentina/Hubert: Update the wiki page http://wiki.eclipse.org/SDMX_Example
with our findings/discussion on SDMX. We've covered a lot of ground
here and I think we need to make sure we capture these thoughts.
2. Martin/Hubert: Update the wiki page http://wiki.eclipse.org/COSMOS_Design_197867
with the new broker content, structure, etc... We need to capture
the alignment with SML concepts, the separation of concerns, query scenarios,
and any API updates.
3. We need to make sure that Jimmy/Joel understand how we reached some
of the conclusions from yesterday's discussion.
4. This will be a significant rev of the broker. I'd like Jimmy/Martin/Bill
to start posting small (<200LOC) patches to the bugzilla. Hubert
and Joel will review them to make sure they comply with the eclipse legal
requirements et. and then begin to check them in.
5. Joel/Hubert: We should target Aug 31 for integration of the new broker
into the code.
6. Balan: Sept 7 for a new rev of the end to end with this new broker code.
We'll need to review the impact to Sheldon.
Does this plan for the next few weeks make sense?
-mw
_______________________________________________________________________________________________________________
Mark Weitzel | STSM | IBM Software Group | Tivoli | Autonomic Computing
| (919) 543 0625 | weitzelm@xxxxxxxxxx
"Mohsin, Jimmy"
<Jimmy.Mohsin@xxxxxx>
08/21/07 02:48 AM
|
To
| "Simmonds, Martin"
<Martin.Simmonds@xxxxxx>, "Valentina Popescu" <popescu@xxxxxxxxxx>,
Mark D Weitzel/Raleigh/IBM@IBMUS
|
cc
| "Cosmos Dev" <cosmos-dev@xxxxxxxxxxx>
|
Subject
| RE: COSMOS DC: Review of Indexing examples |
|
Mark,
1. Agree with Martin on his comment “with
the structure of keysets and dataset types is where I see this going”
2. I agree 100000% percent with Valentina’s statement “SDMX
is meant to be used for exchanging statistical data and any other usage
of it is accidental and not meant to from a spec”.
I think for a VERY long time, SDMX was seen as THE mechanism for
ALL our use cases. As I have stated in past meetings numerous
times, we have MANY high-value use cases that shuttle a VERY small amount
of data around. Using SDMX to fulfill these cases would have made
no sense whatsoever.
3. That said, we should certainly use SDMX for what is meant to be used
for, i.e. for managing statistical data. We have a FEW use cases
in this space as well. However, if I had my way, I would implement
some SIMPLE use cases first where I would demonstrate the simpler (non-SDMX)
use cases first….
4. Type support based registration makes a LOT of sense…
Thanks,
Jimmy Mohsin
+1-609-635-1703
From: Simmonds, Martin
Sent: Tuesday, August 21, 2007 1:59 AM
To: Valentina Popescu; Mark D Weitzel
Cc: Mohsin, Jimmy
Subject: RE: COSMOS DC: Review of Indexing examples
I agree that the Data Broker registration based on SML/CML is a better
alternative. I don’t want to see SDMX mentioned in all of this,
pure xml with the structure of keysets and dataset types is where I see
this going.
I am going to show Bill Muldoon the Cosmos Web Gui later today, and I think
that by the end of today we can agree on what is in the Data Broker.
I had asked a few weeks ago for someone to go over SML. Is it possible
that we could go over a scenario where SML is used? This will speed
up things if we have a good understanding. As yet I don’t see where
CA’s use cases would use SML, I do most definitely see how CML should
be used for the Dataset Type. That said, how do you ensure we are
using CML? Is this an I6 possibility or after?
Martin...
From: Valentina Popescu [mailto:popescu@xxxxxxxxxx]
Sent: 21 August 2007 05:12
To: Mark D Weitzel
Cc: Mohsin, Jimmy; Simmonds, Martin
Subject: RE: COSMOS DC: Review of Indexing examples
Sorry I couldn't make today's meeting, I have a full day event for Mon-Wend
this week
I have to agree with Mark on the SDMX proposal. I had the to-do for mapping
SDMX to the SML Data Center types so I investigated more the SDMX specifications
and looked at SDMX samples and proposed approaches
My conclusion
- SDMX is meant to be used for exchanging statistical data and any other
usage of it is accidental and not meant to from a spec definition
- If the only purpose of using SDMX with the Data Broker is to make use
of the keysets support and data-type separation, than we just complicate
unnecessarily the entire usecase. There are other ways, much cleaner and
easier to get to the same support ( such as pure xml .. )
- since we are building our scenarios around a Data Center solution and
the project goal is to support and align with SML and SML based resources,
I would rather incline to use the Data Center sample in COSMOS for the
type support registration. This also has the benefit that is SML enabled
and will eventually become CML; a Data Broker registration based on SML/CML
is probably something that should be considered as a better alternative
Thank you,
Valentina Popescu
IBM Toronto Labs
Phone: (905)413-2412 (tie-line 969)
Fax: (905) 413-4850
Mark D Weitzel/Raleigh/IBM@IBMUS
08/20/2007 09:21 AM
|
To
| "Simmonds, Martin" <Martin.Simmonds@xxxxxx>
|
cc
| "Mohsin, Jimmy" <Jimmy.Mohsin@xxxxxx>,
Valentina Popescu/Toronto/IBM@IBMCA
|
Subject
| RE: COSMOS DC: Review of Indexing examplesLink |
|
Martin,
My intent is to keep the data broker as simple as possible. I agree
that it's just a "yellow pages" kind of thing.
What I would like to do is keep all the information inside of COSMOS consistent
with itself. Here's what we got right now....
Inside the broker, we have SDMX
Inside our repository, we have SML
It seems that we all agree on the following:
* We don't want to use all of SDMX, only this interesting aspect around
the way data is described.
* We like the fact the kind/type of data is kept distinct from the source
of the data
>From my perspective, what I was trying to do was to determine if there
was a way we could use the constructs that we already have defined in SML
as way to achieve the same thing. The thinking here is that if we
can use the same concepts from SML to have the indexes, then the things
in our repository (ultimately CML models) will be able to line up nicely.
I'm hoping we can simplify the COSMOS design by reducing our dependency
on another specification. We can use today's call to glean the true
requirements on the broker and figure out if this is possible.
BTW... Hubert should have asked you guys to start attaching code to the
enhancement request. We can coordinate with him today, but we should
be able to start with some small patches that we can begin to work together
on.
other comments in line
<mdw>
like this...
</mdw>
_______________________________________________________________________________________________________________
Mark Weitzel | STSM | IBM Software Group | Tivoli | Autonomic Computing
| (919) 543 0625 | weitzelm@xxxxxxxxxx
"Simmonds, Martin"
<Martin.Simmonds@xxxxxx>
08/20/07 08:24 AM
|
To
| Mark D Weitzel/Raleigh/IBM@IBMUS
|
cc
| "Mohsin, Jimmy" <Jimmy.Mohsin@xxxxxx>
|
Subject
| RE: COSMOS DC: Review of Indexing examples |
|
Thanks Mark,
I don’t remember seeing this, it must have got lost in the mail J
Mark, On that last long long meeting I got cut off, and you were starting
to talk about SML and facets, when I got back. I was not sure where
that came from, but I am guessing the following (or maybe
its just another Simmonds idea J)
:
The Data Broker is a yellow pages that stores a KeySet Name (Classification)
and EPR to the DM.
<mdw>
agree
</mdw>
This is how CA sees the use of it, yet others seem to want to use the databroker
as more than just this. They see the Databroker as storing some kind
of metadata that you can query via Xpath from the Client.
<mdw>
I think we've been confusing/blending a number of responsibilities. We
need absolute clarity on what the broker does and what it interacts with.
I want this to be as simple as possible.
</mdw>
I am guessing that the Keyset Name/Epr could be a reference to an ‘entry’
in the SML repository, that could contain this extra metadata.
<mdw>
In my head the "keyset" points to an EPR. The EPR is an
endpoint for an MDR.
That's the only thing the broker does. So i'd write you example as....
"The keyset points to an EPR. The EPR is the CMDBf MDR interface for
the SML repository. CMDBf queries can be used against this MDR interface
to obtain extra metadata.
</mdw>
It would mean that CA uses the DataBroker from a client in the way it has
been , and the other users would get their Client to get all the data managers
via the Data Broker, and to read the metadata from the SML repository keyed
on keyset name and epr.
<mdw>
I'm confused. Can you provide more details on this interaction. Maybe
I'm not sure how CA is using the broker now.
</mdw>
That would mean the Data Manager would have to register with the DataBroker
and puts its Metadata into the SML repository.
<mdw>
No. I think the Data Manager (a.k.a. MDR)
1) Finds the Data Broker it is supposed to use from it's Management Domain.
2) Registers with the Data Broker that it is available.
3) The Data Broker and the Data Manager "talk to each other nicely"
to exchange the necessary metadata about itself (e.g. keysets).
That's it.
(btw... I think steps 2 & 3 are similar to CMDBf registration, but
I don't want to go there yet b/c I'd like to get a simple case working
first....)
</mdw>
The client would then be able to use Xpath to query against the metadata
that it gets out of the SML Repository.
<mdw>
The client can ask the data broker for MDRs (a.k.a. Data Managers) that
manage. This query will be XPath (?). It get's back an EPR
to an MDR.
</mdw>
Is this way off base?
<mdw>
Not sure... I think we are trying to get to the same thing here, but maybe
I've thrown a Yorker.
P.S. "way off base" I believe this is a baseball
_expression_. Being that you are from the UK, I'd expect, either a)
futball (a.k.a. soccer) reference or b) cricket reference.
</mdw>
Martin...
From: Mark D Weitzel [mailto:weitzelm@xxxxxxxxxx]
Sent: 20 August 2007 13:10
To: Simmonds, Martin
Subject: Fw: COSMOS DC: Review of Indexing examples
You should have seen this invite fly by last week...
-mw
_______________________________________________________________________________________________________________
Mark Weitzel | STSM | IBM Software Group | Tivoli | Autonomic Computing
| (919) 543 0625 | weitzelm@xxxxxxxxxx
----- Forwarded by Mark D Weitzel/Raleigh/IBM on 08/20/07 08:09 AM -----
|
Subject
|
COSMOS
DC: Review of Indexing examples |
|
When
|
| Mon 08/20/2007
| 10:00 AM
|
|
|
|
|
| Mon 08/20/2007
| 12:00 PM |
|
Invitees
|
Invited
| The
following invitees have been invited |
Required
(to)
| cosmos-dev@xxxxxxxxxxx
|
| |
|
|
|
Chair
|
Mark
D Weitzel/Raleigh/IBM |
|
Where
|
Location
| Audio
Conference Access Numbers
USA: 1-877-421-0033 or 770-615-1250
Passcode: 267667 |
Reserved
| No
rooms or resources have been reserved |
|
| |
Review the "low hanging fruit" of action items that we can do.
SDMX index Examples
Mapping to SML of SDMX _______________________________________________
cosmos-dev mailing list
cosmos-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cosmos-dev