I’ll second Sudhakar’s points.  What is the next step?  There is an important 
decision we need to make on the right way to do this integration.

Marlon


From: <Pamidighantam>, Sudhakar <[email protected]<mailto:[email protected]>>
Reply-To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Date: Friday, June 12, 2015 at 4:57 PM
To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: Re: Gap Analysis between GridChem Middleware and Airavata

Supun and Dimuthu:

Thanks very much for the very nice comparative analysis. This reminds me of the 
happy  old days when we went about establishing this model.
Also we need to see if we need to use a combined model (part GridChem 
middleware services and part Airavata)  or to implement the
requirements in Airavatafor bringing GridChem to quick production in Airavata 
context. Is there a way to estimate the effort required in either case.

Sudhakar.

On Jun 12, 2015, at 3:31 PM, Supun Nakandala 
<[email protected]<mailto:[email protected]>> wrote:

Hi All,

I did analyse the gridchem middleware service's (GMS) user and project 
organisation (virtual organisation) functionality. My analysis is based on the 
data models and data organization. I have attached the data model relating to 
the above functionality here with. I would like to summarize some of the VO 
features in gridchem here

1. Unlike Airavata GMS maintains more detailed user information. For Airavata 
we don't need to maintain such detailed information as we have external 
userstore which maintains these data and does the authentication and 
authorization work. User in GMS can have one of three privilege levels (PI, 
Admin, User)

2. There is a notion of user preferences in GMS. This is something that we can 
think and decide whether to add or not to Airavata

3. Similar to Airavata GMS also has the notion of Project. But the project 
model in GMS contains more details than Airavata. It has fields like 
sponsorName, projectStatus, service units awarded, service units requested etc..

4. GMS contains job level scheduling configuration information in the Project 
model (e.g CPUs per job, disGB per job). But I think keeping them in the 
project level is not a good idea as different jobs may need different 
configuration. In Airavata we keep those information per Experiment (in new 
models per Process).

5. Similar to Airavata it is possible to add users to projects.

6. GMS treats Applications (Software) and Computational Resources as Resources. 
Unlike in Airavata these resources can be assigned to projects and to users to 
use in a particular project(ternary relation). This way GMS has implemented 
access controlling on Applications and CRs.

7. GMS also has features like blacklisting a particular user to user an 
application(Software), renewal of projects etc.. which we don't have in 
Airavata.

The execution unit data model in GMS is Job (Similar to Experiment in Airavata) 
and is associated with the Project model. Data models relating to Job and other 
related entities are not shown in the attached document.

Thanks

On Fri, Jun 5, 2015 at 9:15 AM, DImuthu Upeksha 
<[email protected]<mailto:[email protected]>> wrote:
Hi All,

Supun and myself are working on comparing API level similarities and 
dissimilarities of GridChem middleware services and Airavata SDK. Current 
analysis can be found from [1]. Currently most of the functionalities of 
GridChem middleware has been mentioned under separate categories.

[1] 
https://docs.google.com/document/d/1_BIbg1Stb6zgyPjaLSYSBpH-BekPb5fsIsq9MnhihYw/edit?usp=sharing<https://urldefense.proofpoint.com/v2/url?u=https-3A__docs.google.com_document_d_1-5FBIbg1Stb6zgyPjaLSYSBpH-2DBekPb5fsIsq9MnhihYw_edit-3Fusp-3Dsharing&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=7_-LbDwTKOoIiO4P4OLfUTX6lSdjys9jh2AJ7sBl9ag&m=_55bC5SPDhLCc_pbSnZiFgfLqFCXZ9aRc_1UE5FWNbY&s=sQKCta1zAT9q2S9zyRZJHQk36yVHOQRtp-6a5qbECkQ&e=>

--
Regards
W.Dimuthu Upeksha
Undergraduate
Department of Computer Science And Engineering
University of Moratuwa, Sri Lanka



--
Thank you
Supun Nakandala
Dept. Computer Science and Engineering
University of Moratuwa
<gridchem_vo.pdf>

Reply via email to