[jira] [Commented] (AIRAVATA-1084) [GSoC] Prototype Airavata Support for Application Scheduling using Ultrscan usecase

["Scott McCaulay (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AIRAVATA-1084?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13941990#comment-13941990 ] 

Scott McCaulay commented on AIRAVATA-1084:
------------------------------------------

I am interested in working on this.  I think that having one (or a small number) of good use cases like this could help visualize the more generally applicable solution that could be incorporated into Airavata Orchestrator. 

Bringing together already available information from the resource side about availability and queue wait predictions, and the kind of detailed application performance information described here, it should be possible to build additional intelligence into the assignment of jobs to resources.  

My understanding is that this project would be to extend the Orchestrator in such a way as to support the Ultrascan project, with an intention of keeping that functionality general enough to apply to additional projects.

> [GSoC] Prototype Airavata Support for Application Scheduling using Ultrscan usecase
> -----------------------------------------------------------------------------------
>
>                 Key: AIRAVATA-1084
>                 URL: https://issues.apache.org/jira/browse/AIRAVATA-1084
>             Project: Airavata
>          Issue Type: New Feature
>            Reporter: Suresh Marru
>              Labels: gsoc2014, mentor
>
> Before a general purpose application specific monitoring could be built into Airavata Orchestrator, it will be good to prototype a one application specific monitoring and then think about how to generalize it. 
> As an example, Ultrascan application team Borries Demeler and Gary Gorbet were willing to provide a use case and access to information in their database to quert for the parameters that influence calculation time, such as number of datapoints, grid resolution, cluster hardware, noise fits, Monte Carlo iterations, etc. Based on these data, this prototype would do a multi-variate performance analysis and create a prediction model for the length of time it would take for a dataset to complete on a given resource. This will have to profile the 2 dimensional spectrum analysis (2DSA) and possibly Genetic Algorithm. 



--
This message was sent by Atlassian JIRA
(v6.2#6252)