[
https://issues.apache.org/jira/browse/AIRAVATA-1636?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14376041#comment-14376041
]
Suresh Marru commented on AIRAVATA-1636:
----------------------------------------
Hi John,
These are very good questions. Let me take a stab at few of them and let others
chime in as well.
* its a good observation that Airavata records all previous executions. Users
need not wait for benchmark and very well should be able to use the recorded
information as one of the data source to make scheduling decision. Yes and its
applicable to both a centralized installation as well as local installations.
The centralized installations also are multi-tentanted so in theory they are
identical to local installations.
* HPC resources by nature are heterogeneous and non-uniform. The resources are
very general purpose and Airavata is only a small subset of usage. As you
noted, there are users using them directly on terminal, or through other
gateway and workflow systems. Your question is start time predictions, and its
available on atleast XSEDE - http://karnak.xsede.org/karnak/index.html So your
solution can take advantage of such data when available.
* I see this idea as more of a run-time prediction. There are folks like Karnak
author Warren Smith who has ongoing work on this area. If your GSoC proposal
gets accepted, we can loop you with him to get an understanding of the area
more. If you think, that will help you prepare a better gsoc proposal, we can
right away put you in touch. I am guessing you may have enough to start the
proposal.
> [GSoC] Benchmark framework to facilitate Airavata Scheduling
> ------------------------------------------------------------
>
> Key: AIRAVATA-1636
> URL: https://issues.apache.org/jira/browse/AIRAVATA-1636
> Project: Airavata
> Issue Type: New Feature
> Reporter: Suresh Marru
> Labels: gsoc, gsoc2015, mentor
>
> Airavata assists science gateways to execute on multiple computational
> resources. To efficiently schedule applications on resources, Airavata needs
> to understand application performance. Applications are typically complex in
> terms of the models and algorithms they support and internally implemented
> optimization of resources available. The hardware provides additional
> variables in this optimization in terms of memory and computing units that
> can be allocated and time restrictions in the form of queue limits.
> Scheduling adds to this complexity by implementing policies toward enabling a
> particular Science domain and/or maximizing the usage of the resources
> itself.
> Airavata can feed data from historical executions and a framework can be
> built to systematically feed to new experiments (based on existing or totally
> newly devised models) executed. The run and timing data then can be codified
> such that the information can be presented to the user if an intelligent
> choice can be made by the user or can be used programmatically by Airavata in
> cases where the user does not or cannot provide such a choice.
> The end goal of this benchmark exercise will be to provide fastest execution
> time possible accounting for constraints available in the gateway to optimize
> its own allocations for all the users in the communities the gateway
> supports.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
