[ogsa-wg] HPC profile questionaire draft

Marty Humphrey humphrey at cs.virginia.edu
Sun Nov 19 08:41:16 CST 2006


Hiro et. Al.,

As the chair of the HPC Profile WG, I'm a little confused and concerned
about this survey.

Requirements gathering, user-community engagement, and a desire to continue
momentum are all excellent goals. I see this survey as attempting to satisfy
these goals, which is great.

But you're asking people to comment on something that isn't finished. The
analogy is that the car is halfway through the assembly line and you're
asking people how the car company should improve the car's handling around
turns. 

As I tried to explain in my dialogue with Ian, the initial "finished
product" was never intended to be just the HPC Base Profile by itself.
Rather, the very first "usable" product is the Base Profile with a small set
of extensions that the working group will (shortly) define. There's no way
that someone who fills out this survey could know what these extensions are,
because there's been no public discussion of them to date. We're not hiding
anything -- rather, a lot of people worked very hard to get the SC2006
interop demo working, and we just haven't had time to tell each other our
very specific thoughts on what works, what is ambiguous, and what extensions
should be pursued. I personally have participated in a number of these
conversations at SC, and we're planning for a WG call the week of Nov 27 to
discuss everyone's experiences (we haven't picked the date/time yet).

Many of the questions on the survey imply that the HPC Profile WG thinks
that the HPC Base Profile is "complete", and that's a misrepresentation. So
I think this survey, at this time, could be counterproductive. 

-- Marty


-----Original Message-----
From: ogsa-wg-bounces at ogf.org [mailto:ogsa-wg-bounces at ogf.org] On Behalf Of
Hiro Kishimoto
Sent: Sunday, November 19, 2006 9:17 AM
To: ogsa-hpcp-wg at ggf.org; ogsa-wg at gridforum.org; ogsa-bes-wg at ggf.org
Subject: Re: [ogsa-wg] HPC profile questionaire draft

Hi all,

Based on the suggested questions and discussions, I've created survey
template for HPC profile SC06 interop.

Please have a look and give your feedback by email.
This is still STRAWMAN. DO NOT ANSWER.

http://www.zoomerang.com/recipient/survey.zgi?p=WEB225VTDXBJCL

Thanks,
----
Hiro Kishimoto

Ian Foster wrote:
> Marty:
> 
> My words "far from adequate" and "tiny" were not fair, I agree. My 
> comments were meant to be constructive, in the manner you expressed them 
> below.
> 
> Ian.
> 
> Marty Humphrey wrote:
>>
>> Ian,
>>
>>  
>>
>> Your points are very good and absolutely valid. Absolutely. Great points.
>>
>>  
>>
>> While you and I might disagree with such words as "far from adequate" 
>> and even "tiny" in your subsequent email, we agree that it's very 
>> important to scope this properly, for the points you articulated.
>>
>> And to clarify: I don't think it's necessarily those people most 
>> closely connected with the effort who might think or claim that the 
>> scope is broader than it should be in reality. (Is that sentence 
>> parsable?) I have found that the people on the WG calls and those 
>> people implementing the spec generally have the proper perspective on 
>> this work, in my opinion. Of course, the issue now regards MY 
>> perspective :^)
>>
>>  
>>
>> Again, I think you make a great point that we have sure that vendors 
>> don't just implement this Base HPC Profile and claim they're done. 
>> We've always thought that a key to the overall HPC Profile will be the 
>> extensions -- making sure that EVERYONE understands this point is 
>> critical for this effort.
>>
>>  
>>
>> -- Marty
>>
>>  
>>
>> ------------------------------------------------------------------------
>>
>> *From:* Ian Foster [mailto:foster at mcs.anl.gov]
>> *Sent:* Friday, November 17, 2006 5:49 PM
>> *To:* Marty Humphrey
>> *Cc:* 'Subramaniam, Ravi'; ogsa-hpcp-wg at ggf.org; ogsa-bes-wg at ggf.org; 
>> ogsa-wg at gridforum.org
>> *Subject:* Re: [OGSA-BES-WG] [ogsa-wg] HPC profile questionaire draft
>>
>>  
>>
>> Marty:
>>
>> Please don't apologize: you should never let a comment from me go 
>> without response (-:
>>
>> I didn't mean to suggest that the HPC Profile should not be passed to 
>> the GGF editor.
>>
>> However, while we should certainly celebrate our success in creating 
>> the profile and getting people to implement it, I think it is 
>> critically important that we be explicit about how tiny a step we have 
>> taken. Otherwise, we risk backlash from the "public" (when they look 
>> at this spec we are making such a fuss over) and also, potentially, a 
>> loss of focus from the vendors who we need to work towards true 
>> interop. For some vendors, having a spec that they can say they are 
>> compliant with--and then extend in nonstandard ways to address 
>> security, staging, etc., etc.--will be more important than continuing 
>> the hard work of specification.We don't want to let that happen.
>>
>> Ian.
>>
>>
>> Marty Humphrey wrote:
>>
>> Ian,
>>
>>  
>>
>> I'm sorry, but I can't let that "far from adequate" go by without 
>> response. For the real business scenarios that we have been pursuing, 
>> I do not believe that the HPC Profile is "far from adequate". This has 
>> been reinforced by our experience talking with people at SC.
>>
>>  
>>
>> Post-thanksgiving, our WG will refine the HPC Profile somewhat, but I 
>> don't believe that we will have a significant effort before pushing it 
>> to the GGF editor.
>>
>>  
>>
>> As you know, its design is one of extensibility and composability. I 
>> believe that the HPC "Base Profile" will be sufficient for many real 
>> business use cases. I also believe that the planned extensions are 
>> important for many situations.
>>
>>  
>>
>> - Marty
>>
>>  
>>
>> ------------------------------------------------------------------------
>>
>> *From:* ogsa-bes-wg-bounces at ogf.org 
>> <mailto:ogsa-bes-wg-bounces at ogf.org> 
>> [mailto:ogsa-bes-wg-bounces at ogf.org] *On Behalf Of *Ian Foster
>> *Sent:* Friday, November 17, 2006 3:26 PM
>> *To:* Subramaniam, Ravi
>> *Cc:* ogsa-hpcp-wg at ggf.org <mailto:ogsa-hpcp-wg at ggf.org>; 
>> ogsa-bes-wg at ggf.org <mailto:ogsa-bes-wg at ggf.org>; 
>> ogsa-wg at gridforum.org <mailto:ogsa-wg at gridforum.org>
>> *Subject:* Re: [OGSA-BES-WG] [ogsa-wg] HPC profile questionaire draft
>>
>>  
>>
>> I want to understand what people think we need for real interop. The 
>> HPC Profile as currently defined seems far from adequate for that 
>> purpose. We've mention staging and security as two issues. There are 
>> presumably others.
>>
>> Ian.
>>
>>
>> Subramaniam, Ravi wrote:
>>
>> Hi Andrew,
>>
>>  
>>
>> One question could be "Is the profile as it stands complete from your 
>> perspective? At least for the focus/business areas or types of 
>> customers that you plan to implement for?"
>>
>>  
>>
>> The motivation is that depending on the verticals (e.g. financial, 
>> pharma) there may be some issue that may need to be resolved to be 
>> successful. In this way at least some target constituencies that can 
>> best benefit from what is currently there can be determined. We can 
>> also begin more targeted campaign at the OGF level towards 
>> consumers/customers/end-users and providers in that segments. Building 
>> traction in one of more specific segments may help build momentum for 
>> others.
>>
>>  
>>
>> Ravi
>>
>> ------------------------------------------------------------------------
>>
>> *From:* ogsa-wg-bounces at ogf.org <mailto:ogsa-wg-bounces at ogf.org> 
>> [mailto:ogsa-wg-bounces at ogf.org] *On Behalf Of *Maguire_Tom at emc.com 
>> <mailto:Maguire_Tom at emc.com>
>> *Sent:* Friday, November 17, 2006 11:12 AM
>> *To:* grimshaw at cs.virginia.edu <mailto:grimshaw at cs.virginia.edu>; 
>> ogsa-wg at gridforum.org <mailto:ogsa-wg at gridforum.org>; 
>> ogsa-hpcp-wg at ggf.org <mailto:ogsa-hpcp-wg at ggf.org>; 
>> ogsa-bes-wg at ggf.org <mailto:ogsa-bes-wg at ggf.org>
>> *Subject:* Re: [ogsa-wg] HPC profile questionaire draft
>>
>>  
>>
>> Is there enough specificity in the profile for your engineers to 
>> successfully implement?  If not what do you need?
>>
>>  
>>
>>  
>>
>> Tom
>>
>> _______________________________________________
>> Tom Maguire
>> +1(845) 729-4806
>>
>> ------------------------------------------------------------------------
>>
>> *From:* ogsa-wg-bounces at ogf.org <mailto:ogsa-wg-bounces at ogf.org> 
>> [mailto:ogsa-wg-bounces at ogf.org] *On Behalf Of *Andrew Grimshaw
>> *Sent:* Friday, November 17, 2006 10:21 AM
>> *To:* 'Ogsa-Wg'; ogsa-hpcp-wg at ggf.org <mailto:ogsa-hpcp-wg at ggf.org>; 
>> ogsa-bes-wg at ggf.org <mailto:ogsa-bes-wg at ggf.org>
>> *Subject:* [ogsa-wg] HPC profile questionaire draft
>>
>>  
>>
>> All,
>>
>> At the OGSA F2F I was tasked with coming up with a series of questions 
>> about the HPC profile experiences to go into a zoomerang, so that an 
>> experiences paper could be generated -- either for HPC profile, JSDL, 
>> and/or BES.
>>
>>  
>>
>> Here is a start. Note -- comment on/add questions. Do NOT answer 
>> them!!! That will come later.
>>
>>  
>>
>> 1)    Will the HPC-Profile be useful to your organization/company? 
>> Explain.
>>
>> 2)    Will you engineer your products to be hpc-profile compliant?
>>
>> 3)    In your opinion, are there any unresolved ambiguities? Explain?
>>
>> 4)    Thoughts on the next profile? Anything in particular from BES or 
>> JSDL (or other specification) that you would like to include.
>>
>> 5)    Approximately how many person-days did it take you to implement 
>> the HPC-profile?
>>
>> 6)    What would you change in the HPC-profile? Why?
>>
>>  
>>
>> Remember this is a STRAWMAN ONLY for discussion! Please add, suggest 
>> removal, or modify the above.
>>
>>  
>>
>> A
>>
>>  
>>
>> Andrew Grimshaw
>>
>> Professor of Computer Science
>>
>> University of Virginia
>>
>> 434-982-2204
>>
>> grimshaw at cs.virginia.edu <mailto:grimshaw at cs.virginia.edu>
>>
>>  
>>
>>  
>>  
>>  
>>
>>
>> ------------------------------------------------------------------------
>>
>>
>>  
>>  
>>  
>> -- 
>>   ogsa-wg mailing list
>>   ogsa-wg at ogf.org <mailto:ogsa-wg at ogf.org>
>>   http://www.ogf.org/mailman/listinfo/ogsa-wg
>>
>>
>>
>>
>>
>> -- 
>>  
>>    Ian Foster, Director, Computation Institute
>> Argonne National Laboratory & University of Chicago
>> Argonne: MCS/221, 9700 S. Cass Ave, Argonne, IL 60439
>> Chicago: Rm 405, 5640 S. Ellis Ave, Chicago, IL 60637
>> Tel: +1 630 252 4619.  Web: www.ci.uchicago.edu 
>> <http://www.ci.uchicago.edu>.
>>       Globus Alliance: www.globus.org <http://www.globus.org>.
>>
>>
>>
>>
>> -- 
>>  
>>    Ian Foster, Director, Computation Institute
>> Argonne National Laboratory & University of Chicago
>> Argonne: MCS/221, 9700 S. Cass Ave, Argonne, IL 60439
>> Chicago: Rm 405, 5640 S. Ellis Ave, Chicago, IL 60637
>> Tel: +1 630 252 4619.  Web: www.ci.uchicago.edu 
>> <http://www.ci.uchicago.edu>.
>>       Globus Alliance: www.globus.org <http://www.globus.org>.
> 
> 
> 
> ------------------------------------------------------------------------
> 
> --
>   ogsa-hpcp-wg mailing list
>   ogsa-hpcp-wg at ogf.org
>   http://www.ogf.org/mailman/listinfo/ogsa-hpcp-wg

--
  ogsa-wg mailing list
  ogsa-wg at ogf.org
  http://www.ogf.org/mailman/listinfo/ogsa-wg



More information about the ogsa-wg mailing list