NAREGI PSE material for F2F meeting
Folks, Here is a material for the following agenda. We welcome your questions or comments. Regards, -- Motohiro Yamada
On Mon, 23 May 2005 09:55:25 +0900, kfukui@labs.fujitsu.com (Keisuke Fukui) wrote:
*** DAY 1 (May 23) *** (snip) 16:45 Review of the NAREGI PSE information. (KF; 30 min) - Material will be posted on the acs-wg by noon EDT. - Review the information if there are questions or issues. - Return comments or questions to PSE team to prepare the call in the next morning.
Folks, At page 6 in the slides I sent,
Execution 2. PSE submits a job to Super-Scheduler
should be "2. Submit a job to Super-Scheduler using WFT". NAREGI PSE does NOT submit a user job. Attached is a revised material, please use it for the discussion. Regards, -- Motohiro Yamada
On Mon, 23 May 2005 21:23:57 +0900, moto.yamada@jp.fujitsu.com (Motohiro YAMADA) wrote:
[1 <text/plain; US-ASCII (7bit)>] Folks,
Here is a material for the following agenda. We welcome your questions or comments.
Regards, -- Motohiro Yamada
On Mon, 23 May 2005 09:55:25 +0900, kfukui@labs.fujitsu.com (Keisuke Fukui) wrote:
*** DAY 1 (May 23) *** (snip) 16:45 Review of the NAREGI PSE information. (KF; 30 min) - Material will be posted on the acs-wg by noon EDT. - Review the information if there are questions or issues. - Return comments or questions to PSE team to prepare the call in the next morning. [2 naregi-pse-scenario-20050524.pdf <application/octet-stream (base64)>]
Yamada-san and PSE team, Thank you for sending the material about NAREGI PSE. We reviewed it today and had some questions and a comment. Could you please have those covered in the call for DAY2? Note: we changed the dial in number as Pete announced. Date & Time: May 24 Wed 8:00 EDT/May 24 Thu 21:00 JST Dial in numbers: 1-877-421-0023 1-770-615-1382 Conference code/pin: 252845 See you then! -Keisuke 1. Our understanding of the slides. We understand the slides as below: Slide 3: we see no difference from previous presentation in March, except for the "exec. for trial" was changed to "test run". Slide 5: No change. Slide 7,8: We think this is new and is about a design of Application Pool. We had a difficulty understand these but we think we finally understand these with a background situation: For the coupled simulations, they may share a part of the applications that compose a simulation. Let's say we have three applications A, B and C. Simulation S1 use app A and B. Simulation S2 use app B and C. For such cases, separate archives will be created for each of the applications. Having such separate archives make it possible for the coupled simulations share the common application B. This our understanding as a background for Slide 7 and 8. 2. Questions to NAREGI PSE presentation. - Is our understandings above correct? - What does the blue box mean in slide 7. - In slide 7. Does the deployment information come from ACS? - In slide 7, What does Information Service contain? - In slide 8. Explain the below sentence. "NAREGI PSE will manage them outside of Application Repository." - Isn't it possible that single archive contains hierarchical applications and their relationship? IUDD can. ACS can contain that. CDDLM can deploy that. Additional work flow and/or controlling application will make them work. Comments. - We understood that PSE creates as many archive as source code and compose a coupled simulation from multiple of those archives. It is doable using ACS, but we don't think this take full advantage of the integration capability of ACS.
Keisuke and all, Thank you for your discussion and comments. Here is a response to your questions.
On Tue, 24 May 2005 11:49:55 +0900, kfukui@labs.fujitsu.com (Keisuke Fukui) wrote:
2. Questions to NAREGI PSE presentation. - Is our understandings above correct?
Yes, it is.
- What does the blue box mean in slide 7.
The blue boxes mean OGSA related services covered by NAREGI PSE. But they have no deep meaning.
- In slide 7. Does the deployment information come from ACS?
No. The deployment information means which machine has the deployed application. The deployment information comes from the application information via deployment client in user-level service.
- In slide 7, What does Information Service contain?
Mainly, Information Service provides resource information such as a kind of cpu, os, available memory size and software resources. FYI, NAREGI Information Service uses DMTF-CIM (Common Information Model) for managing the resource information.
- In slide 8. Explain the below sentence. "NAREGI PSE will manage them outside of Application Repository."
This means that an application archive contains one specific target. We are considering the relationship AA for source files and AA for the binary will be managed in the Applicaiton Information on left side of ACS box. I think it is relating to the comments below.
- Isn't it possible that single archive contains hierarchical applications and their relationship? IUDD can. ACS can contain that. CDDLM can deploy that. Additional work flow and/or controlling application will make them work.
Yes, of course. As we do not have a lot of information about the capability of ACS, we have considered based on our current Application Pool at present. We can continue to discuss about it. But about applying CDDLM to our deployment service, we have the feeling that CDDLM is too complicated for our deployment requirements, since our deployment service is focusing on a legacy application and has completed before submitting a job. So, we also need to consider our deployment service.
Comments. - We understood that PSE creates as many archive as source code and compose a coupled simulation from multiple of those archives. It is doable using ACS, but we don't think this take full advantage of the integration capability of ACS.
As I described above, we welcome to discuss about it. Would you give us the detailed information relating to the capability of ACS later? Regards, -- Motohiro Yamada
participants (2)
-
Keisuke Fukui -
Motohiro YAMADA