[dcifed-wg] DCIFed status

[Gary Mazz]

Hi Al,

Comments In line...

On 6/7/2011 3:46 AM, alexander.papaspyrou at tu-dortmund.de wrote:
> Hey Gary,
>
> Am 30.05.2011 um 22:33 schrieb Gary Mazz:
>
> --8<-- snip --8<-- 
>
>> Back to DCI-Fed
>>
>> We can take a few strategies for use cases:
>>
>> 1) Develop our own...
> Nope. I don't think use cases should come from within the working group if they are not endorsed by any external project. That said, all group members (also those "only" on the mailing list) are more than welcome to submit use cases, even if they are simple. DCIfed is in the stage of trying to find focus.
I think there are two ways of looking at internally developed use cases
1) Create a Use Case(s) to support the overall vision of the project, 
where other use cases fall short
2) Create  Use Case that will be able to create new opportunities for 
DCI-Fed adopter, even if no one is doing them today.

This is where innovation is interjected into the process, otherwise we 
end up just rehashing old models

>> 2) Leverage use cases referenced in other intiatives and stds we are considering for inclusion in DCI-Fed
> IIRC, the following points are floating around for a while now:
>
> - the original "DGSI" use case (federation of traditional, community-focused Grid environments)
Agreed
> - the "EGI" use case with OCCI + OVF + CDMI (maybe something else here?) + Federated Security
Maybe I'm missing something ?? I don't see the value proposition at this 
level, maybe OCCI and CDMI at  the next level of detail.
> - the "Gary" use case (see other mails)
These are not "my" use cases, but requirements of a $40B IT consumer in 
the US Fed
> While DGSI is the most advanced (specs have been written in the project and most problems have been cleared out), it also is the one with the least impact, addressing a very specific problem closely related to research infrastructures.
DGSI has some very good specifications, but is not a clear solution. It 
does require a rather costly integration, tough to justify in the 
current economy.
> The EGI use case, in turn, is the one moving forward fastest and being very close to what has been done in the last months by Thijs and Andy within OCCI anyway. Alas, I am not aware how much interest comes from Steven and David to really push this forward in DCIfed (they have published an "EGI" document on this already, see https://www.egi.eu/indico/getFile.py/access?resId=0&materialId=1&confId=415). Regarding the "Gary" use case, we can just discuss this on the list.
I like this EGI document. It the first that ca be described as a true 
reference architecture for deploying VMs in the Grid.  I like the 
concept of the message bus, but it need to be broken out in more detail. 
The editorial mentions the bus may be a series of brokers, which makes 
them VOs. I believe part of DCIfed is to provide composition and 
orchestration between VOs participating in on the message EGI message 
bus. Maybe a service composition model, ala TINA/CORBA/SOA.

EGI is only one use case. We need to also look at the overall 
requirements relevant for cloud computing.
> With respect to Craig's proposal to compile use cases, I fully agree. I'd therefore propose that we collect the "DCIfed internal" use cases (i.e. the ones that have been submitted to the group already or will be submitted in the future), and align them with the ones mentioned by Craig.
Agreed. I posted the US Fed requirements to kick start the work... as 
the first entries to the list. I'm compiling a list of other 
specifications that may prove valuable to the possible approaches we can 
take.

> Taking up Vivek's document management workflow is surely a good idea; we'd still need someone to endorse it in the group, though.
>
The document management work flow is not as easy as one would like to 
believe, especially for the US Fed. There are several types of document 
management work flows deployed in the Fed. Some are mandated by law 
while other are entrenched in enterprise architectures, none easily 
changed. The other issue with the document management work flow is 
different agencies have individual document managed states and work 
flows, each optimized for the organization's mission. Some work flows 
are extra-organizational, between agencies and departments.

What is needed is a scheme for managing documents by separating the 
document artifact from the work flow process. Then artifacts can be 
shared across multiple work flows with each work flow maintaining its 
document state applicable for the appropriate mission.

This case is why composition and orchestration is incredibly important 
for DCIfed. (IMO)
>> 3) Refer to documented best practice white papers and case studies.
> I'd rather use this for reference purposes in any upcoming document.
Agreed-- and the intent
>> Before we jump in with two feet and having all oars driving in a direction, we need to have a meeting on vision and outcome, just to make sure we are all on the same page.
> Agreed. I will schedule a session for the upcoming OGF in Salt Lake City in mid of July. It's co-located with TeraGrid this year, so there should be enough reason to come.
I'll try to make  Salt Lake. Funding is becoming an issue again.
> Best,
> Alexander
>
>> On 5/30/2011 11:46 AM, Craig Lee wrote:
>>> Alex, Gary, et al.,
>>>
>>> I strongly encourage a short write-up on OCCI/CDMI/OVF.  These three standards were mentioned as a group multiple times at the NIST Cloud Standards Summit (both Alan and I attended).  This group also appears in the NIST Cloud Standards Roadmap -- the first complete draft should be available this June.  I can't promise this material would get into the Roadmap, but we need to make the strongest possible, defensible recommendations wrt existing and emerging standards for cloud computing.
>>>
>>> Wrt Use Cases -- and broad use case requirements -- many sets of cloud use cases have been compiled, even among the NIST Cloud WGs.  If DCIFed wants to compile another set of use cases, I would recommend taking some time to at least review and comment on the other sets that have been compiled.  I agree with the use case requirements that Gary mentions in his earlier email.  I would add, though, that Vivek Kundra has said that _document management_, i.e., document workflow, is the single largest use case for the US gov.  (This is probably true for all governments!) It also seems to me that secure access through mobile devices, i.e., smartphones, is also high on the list.
>>>
>>> --Craig
>>>
>>> On 5/30/11 7:48 AM, alexander.papaspyrou at tu-dortmund.de wrote:
>>>> Hey Gary,
>>>>
>>>> putting your last message to the list as well...
>>>>
>>>> Am 23.05.2011 um 01:00 schrieb Gary Mazz:
>>>>
>>>>> The other cloud groups are trying to coordinate.... We will need to leverage this work
>>>> Most definitely.
>>>>
>>>>> NIST is looking to compiles a list standards, use cases and scenarios  that can be applied to cloud and distributed computing. We will need to leverage this work also.
>>>> Sure. Maybe we can do a short write up of how the OCCI/CDMI/OVF integration would address the NIST stuff. I remember Alan pointing me to some NIST standards wiki pages with a table of use cases and cloud desiderata -- would be a good starting point to show what the DCIfed integration is supposed to address, if we assume OCCI/OVF/CDMI to be this integration.
>>>>
>>>>> I'm trying to put a cloud consumer working group together around DCI-Fed. Tony DiCenzo Director Standards committed Oracle's assistance. CA may also participate, if we ask nice. :) Use case and scenario work needs to proceed with the user community
>>>> That sounds very interesting. Can we help you with anything on this?
>>>>
>>>>> I also just sent an email on "hot issues" from the largest IT consumer in the US Fed.., Its part of defining requirement areas for technical focus.
>>>> Yep, seen that. Comments in that mail.
>>>>
>>>> Best,
>>>> Alexander
>>>> _______________________________________________
>>>> dcifed-wg mailing list
>>>> dcifed-wg at ogf.org
>>>> http://www.ogf.org/mailman/listinfo/dcifed-wg