For the WG draft of the ACS Spec.
Folks, Even though we are waiting for the minutes of GGF13 ACS-WG sessions, I'd like to summarize our achievement at GGF13 and start discussion toward the creating the WG draft of the ACS specification by the next GGF14 in June. Through out the two sessions at GGF13, we successfully demonstrated demands and expectations on ACS. The NAREGI PSE, introduced at the first session, is among the promising embodiments of the ACS. We also saw the larger vision lying beyond the initial ACS spec. from the Pete's slides. We then proceed to create the WG draft of the spec. by the GGF14. To move this ahead, I am creating a note to start our activity hoping to post here by the end of tomorrow. I welcome your opinions and comments. If you have plans regarding to our activity, please feel free to express those and let us discuss. I appreciate your contributions to our dream. -Keisuke
Folks, As I announced I propose agenda for our discussions. I believe we need to visit a number of issues and considerations prior to have a WG draft of the ACS specification, which includes: - the new requirements from NAREGI PSE, - relationship to other GGF WGs, such as CDDLM-WG and planned WGs for Execution Management System, - collaboration with IUDD standardization group and - recruiting involvement from the European communities. I will describe them in a bit more detail below: a) the new requirements from NAREGI PSE - including source code to be compiled for each target platform at or prior to the deployment of the application. - storing compiled binaries for target platforms in the repository. - storing (caching) the input and output data pairs through the lifecycle of the multiple executions of the application. b) relationship to the deployment API and component mode being discussed in the CDDLM-WG. - 'addFile' interface in the deployment API looks to be a candidate contact point with ACS, for example, "register" interface of ARI. - where the ACS interfaces should stand in the CDDLM's lifecycle model. - other (expected) GGF WGs such as OGSA design teams. c) collaboration with IUDD standardization group - Archive standard will be output from IUDD standardization. How can we contribute to it, to archive the common format for the grid deployment. Probably we should discuss Grid specific consideration at ACS, then report our outcome to the IUDD group. - Repository interface standard for Grid system will be output from GGF ACS. d) recruiting involvement from the European communities. One of the comments I got after the sessions is that we should seek for the participants from European Grid communities since they share much with our usecase and scope. We need a research on this possibility. UniGrid, EGEE, omii, and more... My plan is having a brainstorming on all of these agenda. Originally, it was planned to discuss about our plan including the above at ACS-WG#2 session. It turned out, however, more difficult to perform this at the presence of the audience explaining what is being discussed and with the limit of time. To do the brainstorming the items, we have options below: - Tele-conferences several times, one or two hours each. - Face-to-face meeting probably one or two days. At the U.S. East coast or in Japan. We can do either or both depending on our schedule, in addition to the regular mailing list discussion. Since we have not discussed the items at the GGF13, I hope we discuss on this as soon as possible. Can we have a tele-conference to talk about these next week? My proposal for the time is 3/30 Wed 19:00EST (3/31 Thu 9:00JST). Please respond if this works for you. Also, please feel free to add agenda and give your feedback to the above. I hope that we are starting a great work from here. -Keisuke
Agenda and time is good for me, assuming the hotel I'll be at has a good connection. Will we be using Skype or what would be the dial-in #? Also, I would have to leave the OGSA teleconference a little early, which would be okay, I think. Keisuke Fukui wrote:
Folks,
As I announced I propose agenda for our discussions.
I believe we need to visit a number of issues and considerations prior to have a WG draft of the ACS specification, which includes:
- the new requirements from NAREGI PSE, - relationship to other GGF WGs, such as CDDLM-WG and planned WGs for Execution Management System, - collaboration with IUDD standardization group and - recruiting involvement from the European communities.
I will describe them in a bit more detail below:
a) the new requirements from NAREGI PSE - including source code to be compiled for each target platform at or prior to the deployment of the application. - storing compiled binaries for target platforms in the repository. - storing (caching) the input and output data pairs through the lifecycle of the multiple executions of the application. b) relationship to the deployment API and component mode being discussed in the CDDLM-WG. - 'addFile' interface in the deployment API looks to be a candidate contact point with ACS, for example, "register" interface of ARI. - where the ACS interfaces should stand in the CDDLM's lifecycle model. - other (expected) GGF WGs such as OGSA design teams. c) collaboration with IUDD standardization group - Archive standard will be output from IUDD standardization. How can we contribute to it, to archive the common format for the grid deployment. Probably we should discuss Grid specific consideration at ACS, then report our outcome to the IUDD group. - Repository interface standard for Grid system will be output from GGF ACS. d) recruiting involvement from the European communities. One of the comments I got after the sessions is that we should seek for the participants from European Grid communities since they share much with our usecase and scope. We need a research on this possibility. UniGrid, EGEE, omii, and more...
My plan is having a brainstorming on all of these agenda. Originally, it was planned to discuss about our plan including the above at ACS-WG#2 session. It turned out, however, more difficult to perform this at the presence of the audience explaining what is being discussed and with the limit of time. To do the brainstorming the items, we have options below:
- Tele-conferences several times, one or two hours each. - Face-to-face meeting probably one or two days. At the U.S. East coast or in Japan.
We can do either or both depending on our schedule, in addition to the regular mailing list discussion. Since we have not discussed the items at the GGF13, I hope we discuss on this as soon as possible. Can we have a tele-conference to talk about these next week? My proposal for the time is 3/30 Wed 19:00EST (3/31 Thu 9:00JST). Please respond if this works for you.
Also, please feel free to add agenda and give your feedback to the above. I hope that we are starting a great work from here.
-Keisuke
Happy Easter! -- Michael Behrens R2AD, LLC (571) 594-3008 (cell) *new* (703) 714-0442 (land)
FYI, I have pre-scheduled a tele-conference. If you prefer another time, please let us know about it. * Conference Details Date and Time: 3/30 Wed 19:00EST = 3/31 Thu 9:00JST Dial-in Number: 1-831-600-1000 (Santa Cruz, California) Participant Access Code: 7771111 -Keisuke Keisuke Fukui wrote:
I hope we discuss on this as soon as possible. Can we have a tele-conference to talk about these next week? My proposal for the time is 3/30 Wed 19:00EST (3/31 Thu 9:00JST). Please respond if this works for you.
Folks, My apologies that I got the time conflict with my schedule for the next tele-conference schedule that we had talked in the call. May I suggest an alternative? Next call: Apr 20 Wed 20:00 EST/Apr 21 Thu 10:00 JST If anyone has problems with this, please feel free to respond to this message. -Keisuke Keisuke Fukui wrote:
FYI,
I have pre-scheduled a tele-conference. If you prefer another time, please let us know about it.
* Conference Details Date and Time: 3/30 Wed 19:00EST = 3/31 Thu 9:00JST Dial-in Number: 1-831-600-1000 (Santa Cruz, California) Participant Access Code: 7771111
-Keisuke
Keisuke Fukui wrote:
I hope we discuss on this as soon as possible. Can we have a tele-conference to talk about these next week? My proposal for the time is 3/30 Wed 19:00EST (3/31 Thu 9:00JST). Please respond if this works for you.
Just re-read the latest draft ACS specification and wanted to send in some more feedback - looking good! - Recommend that the requirements be amended to include the compilation model as discussed in the last telecom. - Are there two types of dependencies: install/deploy time and runtime? Perhaps the deploy time could include the compilation engines whereas the runtime could just include what is needed to run the job. Just a thought as perhaps based on this it might be possible for clients of ACS to inquire what is needed ahead of time to prepare, etc. - Should AAID be defined using the results of WS-Naming working group? - Security - Would the signature be created and checked automatically by the implementation of ARI or by Business Activity Managers? - Is there a possible relationship between ACS and software such as Install Shield or Install Anywhere? Curious. Looking forward to the next telecom on the 20th! -- Michael Behrens R2AD, LLC (571) 594-3008 (cell) *new* (703) 714-0442 (land)
Hi Mike, Michael Behrens wrote:
Just re-read the latest draft ACS specification and wanted to send in some more feedback - looking good!
Thanks. Here are some comments on your questions.
- Recommend that the requirements be amended to include the compilation model as discussed in the last telecom.
Yes, we should discuss more in detail; it sounds reasonable to facilitate this kind of use case at this moment, though I guess we need a discrimination how much we can increment the archive in the repository in terms of size and variation.
- Are there two types of dependencies: install/deploy time and runtime? Perhaps the deploy time could include the compilation engines whereas the runtime could just include what is needed to run the job. Just a thought as perhaps based on this it might be possible for clients of ACS to inquire what is needed ahead of time to prepare, etc.
I understand as you pointed. I'd like to hear more from Naregi engineers on this point.
- Should AAID be defined using the results of WS-Naming working group?
I guess so, and we need to find out how IUDD address this.
- Security - Would the signature be created and checked automatically by the implementation of ARI or by Business Activity Managers?
I believe so.
- Is there a possible relationship between ACS and software such as Install Shield or Install Anywhere? Curious.
Install Shield like products handles more generic software installation rather than the grid applications. In my opinion, they are more relevant with IUDD. AAF of the ACS itself should be positioned as a grid extension of the generic software installation which is the IUDD. One question is that AAF can be an additional specification to the existing IUDD or it is an addition AND subtraction to the IUDD specification, to be a grid extension. -Keisuke
Folks, I ask for a change of the call schedule after consulting with Tom. Apr 12 Tue 19:00 EDT/Apr 13 Wed 8:00 JST If this doesn't work for you, please respond to this e-mail ASAP. I have tentatively reserved a call bridge: Dial-in Number: 1-702-851-3330 (Las Vegas, Nevada) Participant Access Code: 7771111 *** The number is different from the usual one *** -Keisuke Keisuke Fukui wrote:
Folks,
My apologies that I got the time conflict with my schedule for the next tele-conference schedule that we had talked in the call. May I suggest an alternative?
Next call: Apr 20 Wed 20:00 EST/Apr 21 Thu 10:00 JST
If anyone has problems with this, please feel free to respond to this message.
-Keisuke
Keisuke Fukui wrote:
FYI,
I have pre-scheduled a tele-conference. If you prefer another time, please let us know about it.
* Conference Details Date and Time: 3/30 Wed 19:00EST = 3/31 Thu 9:00JST Dial-in Number: 1-831-600-1000 (Santa Cruz, California) Participant Access Code: 7771111
-Keisuke
Keisuke Fukui wrote:
I hope we discuss on this as soon as possible. Can we have a tele-conference to talk about these next week? My proposal for the time is 3/30 Wed 19:00EST (3/31 Thu 9:00JST). Please respond if this works for you.
Folks, Here is an info. for the call tomorrow. See you there! -Keisuke ### Logistics for the call Date & Time: Apr 12 Tue 19:00 EDT/Apr 13 Wed 8:00 JST Dial-in Number: 1-702-851-3330 (Las Vegas, Nevada) Participant Access Code: 7771111 *** The number is different from the usual one *** ### Proposed Agenda 1. Current status of the potential F2F participants: Tom: Not available in April. Mike: In Washington D.C., April or May, except for middle of April. Pete: (same as Mike?) Keisuke: April or May, except for the first week of May, US or Japan. Sachiko: April or May, except for the first week of May, US or Japan. Naregi PSE team cannot be present this time in the U.S, but I will get their requirements or proposals, possibly with prior meetings to present at F2F. Also, they can join the meeting on the phone. 2. How does the Naregi PSE on-demand compilation model fit to the ACS and CDDLM? It incrementally adds up the compiled binary for each of the target platform to the part of the archive. They could be variable depending on details such as version of the compiling system and operating system even for an family of the platform. It may affect on the consistency of the initial archive. To me, it would be reasonable to have a separate class of the storage for these from the original archive that is thought to be stable and static. If the binaries need to be created and stored in a repository, is the process part of the deployment or execution? I would imagine it is a part of the deployment, which is heavier in load than deployments for the typical application, but still separate from the execution of the program itself. If so, it means we need once store the output of that type of deployment into the repository, and later for the real execution of the application, it needs to be pulled and deployed to the target platform. The process needs to be clarified in terms of the interaction between ACS and CDDLM. 3. Other topics - Security, naming, other install tools, etc... - IUDD standardization status. End of the call info.
participants (2)
-
Keisuke Fukui -
Michael Behrens