modified plan for jobmanagement discussions
Dear All, During the last teleconference a suggestion to speed up the PGI process was made. It was proposed to start weekly phone meetings (on Wednesdays) to discuss the computational related topics. Andrew volunteered to chair the first meeting and provide a summary of the available documents related to computation (http://forge.gridforum.org/sf/docman/do/listDocuments/projects.pgi-wg/docman...). Since then, the PGI co-chairs have been discussing and reconsidering whether this would really be the most efficient approach, i.e. start a discussion around the interpretation of raw drafts. The jobmanagement related drafts are the results of both the pre-OGF work of the ARC/gLite/Unicore teams and the work undertaken in the GIN community group. The drafts in their current status are rather difficult to be interpreted correctly since they were not meant to be for public consumption, they were written as internal working materials for a smaller group. Since then, we never had the time to put these documents in a more readable form. Thus, we feel that it is our responsibility to first better organize our initial drafts, so that people can make better use of their time by reading the improved document rather than trying to "reverse engineer" our ideas from the current poorly written notes. Therefore, we concluded that the PGI group would make a better progress if the proposed modified plan was followed: 1) the initial authors of the drafts are asked to re-work their internal memos, create a consistent write-up of their ideas, prepare a draft about the "production execution service". This document on "production execution service" should describe all the main aspects of a job management service needed in a production environment. Proposed deadline for the document: 31st March 2) schedule the first jobmanagement phone conference one week after the write-up is available: initially on 8th April. This also means that the teleconference called for this Wednesday is not needed. 3) meanwhile, the security discussions should go forward with full speed and hopefully result in an agreed draft. regards, Balazs, Moreno, Morris PGI co-chairs
1) the initial authors of the drafts are asked to re-work their internal memos, create a consistent write-up of their ideas, prepare a draft about the "production execution service". This document on "production execution service" should describe all the main aspects of a job management service needed in a production environment. Proposed deadline for the document: 31st March
What do you mean by a draft of a 'Production Execution Service'? Is this going to start off with a clean sheet or analyse where the current set of specifications works and where they do not work and propose changes? It seems a great shame now that someone has finally volunteered to move forward on this to stop the work... having the call so that everyone can understand what the issues are (i.e. those that weren't at the original CERN meeting) would seem to be worthwhile use of the call. Steven
Steven Newhouse wrote:
What do you mean by a draft of a 'Production Execution Service'? Is this going to start off with a clean sheet or analyse where the current set of specifications works and where they do not work and propose changes?
It seems a great shame now that someone has finally volunteered to move forward on this to stop the work... having the call so that everyone can understand what the issues are (i.e. those that weren't at the original CERN meeting) would seem to be worthwhile use of the call.
It is important to observe that the PGI activity is not going to stall due to this: as Balazs pointed out, the security discussion must proceed, as a lot of work has been done and we are hopefully going to reach an agreement shortly. As far as the job management task is concerned, we are not going to prepare a full proposal for a PGI execution service: this is something which must be done by the whole PGI working group, and hopefully will follow this initial polishment of the Geneva documents. We were asked many times during the initial PGI teleconferences to provide evidence that the current set of specifications and profiles is not adequate to support the needs for our production infrastructures. We feel that it is time to finally provide an appropriate, written answer to these requests in the form of a well-defined set of requirements for a production-quality job management interface. These requirements would then help focusing the effort of the PGI group on the actual problems. We are unlikely to give appropriate answers to those which were not involved in the CERN meeting during the short timeframe of the call, and the likely outcome is that we will be asked to prepare such a document anyway. Your co-chairs, Balazs, Moreno and Morris. -- Moreno Marzolla INFN Sezione di Padova, via Marzolo 8, 35131 PADOVA, Italy EMail: moreno.marzolla@pd.infn.it Phone: +39 049 8277103 WWW : http://www.dsi.unive.it/~marzolla Fax : +39 049 8756233
Hi Moreno, Currently I am still confused, are you (PGI co-chairs) suggesting that we need a completely new spec or just a document that gives an outline of where the currently specifications are deficient?? David On 16/03/2009 13:41, "Moreno Marzolla" <moreno.marzolla@pd.infn.it> wrote:
Steven Newhouse wrote:
What do you mean by a draft of a 'Production Execution Service'? Is this going to start off with a clean sheet or analyse where the current set of specifications works and where they do not work and propose changes?
It seems a great shame now that someone has finally volunteered to move forward on this to stop the work... having the call so that everyone can understand what the issues are (i.e. those that weren't at the original CERN meeting) would seem to be worthwhile use of the call.
It is important to observe that the PGI activity is not going to stall due to this: as Balazs pointed out, the security discussion must proceed, as a lot of work has been done and we are hopefully going to reach an agreement shortly.
As far as the job management task is concerned, we are not going to prepare a full proposal for a PGI execution service: this is something which must be done by the whole PGI working group, and hopefully will follow this initial polishment of the Geneva documents. We were asked many times during the initial PGI teleconferences to provide evidence that the current set of specifications and profiles is not adequate to support the needs for our production infrastructures. We feel that it is time to finally provide an appropriate, written answer to these requests in the form of a well-defined set of requirements for a production-quality job management interface. These requirements would then help focusing the effort of the PGI group on the actual problems. We are unlikely to give appropriate answers to those which were not involved in the CERN meeting during the short timeframe of the call, and the likely outcome is that we will be asked to prepare such a document anyway.
Your co-chairs, Balazs, Moreno and Morris.
David Wallom wrote:
Hi Moreno,
Currently I am still confused, are you (PGI co-chairs) suggesting that we need a completely new spec or just a document that gives an outline of where the currently specifications are deficient??
The starting point is the second one you mention: deciding whether these deficiencies (motivated by actual requirements from production grids infrastructures) can be filled with brand new specifications/profiles, or using a suitable combination of existing specifications/profiles (or a mix of the two approaches) will follow. I have the following three steps in mind: 1) gather the "original" set of requirements which motivated the creation of the PGI WG. These were the requirements implied in the CERN meeting notes, which are hard to understand in their current form; 2) the original set of requirements is integrated with additional inputs coming from the other PGI participants; 3) starting from that, we can decide whether a combination of existing specifications/profiles can satisfy these requirements, or new profiles or brand new specifications are needed instead. Moreno. -- Moreno Marzolla INFN Sezione di Padova, via Marzolo 8, 35131 PADOVA, Italy EMail: moreno.marzolla@pd.infn.it Phone: +39 049 8277103 WWW : http://www.dsi.unive.it/~marzolla Fax : +39 049 8756233
All, My two cents worth. First, I did not think the documents were hard to understand, though the fourth document on the CREAM architecture did not make recommendations rather documented an implementation. I think most of the proposals can be accommodated rather easily within the scope and spirit of the original BES spec. Second, if I were one of the original authors of the documents (which I am not), I would rather discuss the issues before spending the time to write up something formal. That said, if something formal is desired, by all means. Third, I'm not going to lose any sleep over canceling a telecom - as long as we make good progress the following week. So, are we canceled, or are we go. A -----Original Message----- From: pgi-wg-bounces@ogf.org [mailto:pgi-wg-bounces@ogf.org] On Behalf Of Moreno Marzolla Sent: Monday, March 16, 2009 11:34 AM To: David Wallom Cc: pgi-wg@ogf.org Subject: Re: [Pgi-wg] modified plan for jobmanagement discussions David Wallom wrote:
Hi Moreno,
Currently I am still confused, are you (PGI co-chairs) suggesting that we need a completely new spec or just a document that gives an outline of where the currently specifications are deficient??
The starting point is the second one you mention: deciding whether these deficiencies (motivated by actual requirements from production grids infrastructures) can be filled with brand new specifications/profiles, or using a suitable combination of existing specifications/profiles (or a mix of the two approaches) will follow. I have the following three steps in mind: 1) gather the "original" set of requirements which motivated the creation of the PGI WG. These were the requirements implied in the CERN meeting notes, which are hard to understand in their current form; 2) the original set of requirements is integrated with additional inputs coming from the other PGI participants; 3) starting from that, we can decide whether a combination of existing specifications/profiles can satisfy these requirements, or new profiles or brand new specifications are needed instead. Moreno. -- Moreno Marzolla INFN Sezione di Padova, via Marzolo 8, 35131 PADOVA, Italy EMail: moreno.marzolla@pd.infn.it Phone: +39 049 8277103 WWW : http://www.dsi.unive.it/~marzolla Fax : +39 049 8756233 _______________________________________________ Pgi-wg mailing list Pgi-wg@ogf.org http://www.ogf.org/mailman/listinfo/pgi-wg
Second, if I were one of the original authors of the documents (which I am not), I would rather discuss the issues before spending the time to write up something formal. That said, if something formal is desired, by all means.
I have still not seen any good reason why the call should not happen. Steven
We were asked many times during the initial PGI teleconferences to provide evidence that the current set of specifications and profiles is not adequate to support the needs for our production infrastructures.
OK. Its good to hear that this will finally be done in writing.
We feel that it is time to finally provide an appropriate, written answer to these requests in the form of a well-defined set of requirements for a production-quality job management interface.
I would focus on production-capable rather than production-quality. Quality has nothing to do with the interface - its a matter of implementation.
These requirements would then help focusing the effort of the PGI group on the actual problems. We are unlikely to give appropriate answers to those which were not involved in the CERN meeting during the short timeframe of the call, and the likely outcome is that we will be asked to prepare such a document anyway.
And it is also possible that there are other requirements that are not met in the current specifications that _could_ be incorporated into the work of PGI. Again, it would be nice to see an open discussion of these issues. Steven
Hi,
It seems a great shame now that someone has finally volunteered to move forward on this to stop the work... having the call so that everyone can understand what the issues are (i.e. those that weren't at the original CERN meeting) would seem to be worthwhile use of the call.
some non-technical comments: Like I wrote in my earlier comment, I'm not sure this Wednesday initiative is a move *forward* - rather, sideways, but it is difficult for me to judge from the very brief minutes. Let me explain: I was at the famous CERN meeting, and went home quite elated, as I finally saw a group speaking in practical terms and tackling real-life problems, instead of venturing into yet another academic decomposition into spherically symmetric horses. The documents produced back then were pretty raw and in places unintelligible, but promising. I saw PGI as the proper framework to make these documents developing into practical implementations. And I am saddened to see it getting bloated, bogged in formalities and tormented by various split attempts. It is very unfortunate that the last Friday call was first announced as canceled and then went ahead - this prevented many relevant people from taking part in it. I thus would not regard results of last Friday meeting as anything that has wide implications without further clarifications. And on a slightly more technical part, I received no comments on the Use case; shall I upload it to Wiki, and if yes - where? Cheers, Oxana
On Monday 16 March 2009 14:53, Balazs Konya wrote:
Dear All,
During the last teleconference a suggestion to speed up the PGI process was made. It was proposed to start weekly phone meetings (on Wednesdays) to discuss the computational related topics. Andrew volunteered to chair the first meeting and provide a summary of the available documents related to computation (http://forge.gridforum.org/sf/docman/do/listDocuments/projects.pgi-wg/docman...).
Since then, the PGI co-chairs have been discussing and reconsidering whether this would really be the most efficient approach, i.e. start a discussion around the interpretation of raw drafts.
The jobmanagement related drafts are the results of both the pre-OGF work of the ARC/gLite/Unicore teams and the work undertaken in the GIN community group. The drafts in their current status are rather difficult to be interpreted correctly since they were not meant to be for public consumption, they were written as internal working materials for a smaller group.
Since then, we never had the time to put these documents in a more readable form. Thus, we feel that it is our responsibility to first better organize our initial drafts, so that people can make better use of their time by reading the improved document rather than trying to "reverse engineer" our ideas from the current poorly written notes.
Therefore, we concluded that the PGI group would make a better progress if the proposed modified plan was followed:
1) the initial authors of the drafts are asked to re-work their internal memos, create a consistent write-up of their ideas, prepare a draft about the "production execution service". This document on "production execution service" should describe all the main aspects of a job management service needed in a production environment. Proposed deadline for the document: 31st March
Could You please be more specific? Who is working on what? And why do You use plural? There is only single draft on PGI gridforge - Technology Survey. A.K.
2) schedule the first jobmanagement phone conference one week after the write-up is available: initially on 8th April.
This also means that the teleconference called for this Wednesday is not needed.
3) meanwhile, the security discussions should go forward with full speed and hopefully result in an agreed draft.
regards,
Balazs, Moreno, Morris PGI co-chairs _______________________________________________ Pgi-wg mailing list Pgi-wg@ogf.org http://www.ogf.org/mailman/listinfo/pgi-wg
participants (7)
-
Aleksandr Konstantinov -
Andrew Grimshaw -
Balazs Konya -
David Wallom -
Moreno Marzolla -
Oxana Smirnova -
Steven Newhouse