airavata-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Saminda Wijeratne <samin...@gmail.com>
Subject Re: Need a new data model for launched experiments?
Date Sat, 17 May 2014 09:48:53 GMT
+1

I think previously also we made this distinction and tried to solve it.

Experiment Data = Metadata (name/user/etc)
         + Configurations (Scheduling data / input / etc.)
         + Dynamic Configurations (dynamic scheduling data/generated
scripts/etc.)
         + Results (statuses/intermediate or final output/error)

IMO when creating an experiment user would expect only to give metadata and
the configuration. when analyzing a launched experiment a user will be
interested in all the data mentioned in the equation.


On Thu, May 15, 2014 at 12:21 PM, Marlon Pierce <marpierc@iu.edu> wrote:

> Idly wondering...experiments are the most complicated part of the
> Airavata data model since a) they are stateful, and b) they are used by
> other Airavata components. This has led to some of the problems in
> recent Jiras.
>
> Part of the problem, I'll guess, is that we have fields marked as
> "optional" in experimentModel.thrift that are really "required" because
> we need to allow a gateway to create a skeleton experiment that has the
> details filled in later.
>
> How about a new data model class, launchedExperiment, that represents
> the fundamental change from a "created" experiment to one that has been
> scheduled?
>
> Marlon
>
>

Mime
View raw message