Hi Sachith,

Please find my comments inline


On Tue, Apr 8, 2014 at 10:35 AM, Sachith Withana <[email protected]>wrote:

> Hi all,
>
> This is regarding the issue
> Airavata-1126<https://issues.apache.org/jira/browse/AIRAVATA-1126>
> .
>
> I have a few questions regarding the design.
>
> 1. Is the Application Catalog going to be queryable?
>
Currently we do not have an advance usecase which  requires gateway users
needing to query for application definitions or their deployments out of
application catalog. But there might be a scenario which gateway users
might be interested in selecting deployments based on their static
configurations and/or dynamic status. For example "allow execution in an
clustered resource which is currently not that busy or capable of finishing
the job within 48 hours". We see in Ultrascan, users are given the choice
to select the cluster of which they can run their simulation.
On the other hand querying could be very useful for scheduling purposes.
Thus it is useful feature for Airavata core developers and in rare
scenarios gateway developers also.

2. How are you going to handle the multiple deployments of an application?
>     going to be multiple Application catalog for each deployment or one App
> catalog for each deployment describing where the application is deployed?
>
Current offline discussions we had points to having one globus application
catalog shared among gateways. Since we still do not have a specific design
for this, I'll put forward my general thoughts about the application
catalog.

   1. Single deployment of Airavata will have only one application catalog
   shared among any registered gateways in that Airavata instance.
   2. An application catalog will contain multiple applications and each
   application may have multiple deployments defined for it.
   3. Permissions are placed to control which applications are available to
   which gateway (Airavata admin will configure this)
      1. Gateways can list all applications in the application catalog but
      they can only access the ones which they have permission
      2. permission could be defined for different deployments of the same
      applications
      4. Given an application in the application catalog the access data
   and metadata could be considered in to 2 categories which 1) shared 2)
   gateway specific. (its still a question where we save gateway specific data)
   5. Gateways can request deployment details of an application which it
   has permission to.
   6. Airavata admins can add/remove applications and its deployments,
   enable/disable permissions for resources defined in the application catalog
   for gateways, activate/deactivate resources defined in application catalog
   for gateway(s) for a period of time or permanently etc.

I'm working on a mindmap to gather all the data relating to an application
catalog. I will post it to the list once I have a decent looking one.

Question I do have in my mind is where should the resources such as
resource paths defined. Whether it should be part of the application access
data or should be independent and later linked to application access data?
The reason is if we are planning to allow GFac to handle file management
Tasks in future it might be easier to separating out resource paths from
application access data. wdyt?

>
> --
> Thanks,
> Sachith Withana
>

Reply via email to