[
https://issues.apache.org/jira/browse/AIRAVATA-346?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13613936#comment-13613936
]
Shameera Rathnayaka commented on AIRAVATA-346:
----------------------------------------------
Hi devs,
With the knowledge gained from the paper, let me briefly explain how emulation
process works and it states. The paper suggests 9 steps to an emulated process
but i have come up with a 7 steps flow which is sufficient to do our works
according to my knowledge.
Each job submission to application emulator from workflow engine can consider
application service execution flow.
We can divide such a flow into 7 steps which will call state of the flow,
Flow start-->Input Data transfer --> Pre-computation --> Computation -->
Post-computation --> Output data transfer --> Flow complete
Flow start - In this state it processes initialization activity like resource
and flow information gathering.
Input data transfer - Captures the data transfers, this is not required when
inputs are absent.
Pre computation - Input data products registered with meta-data catalog.
Computation - Emulates the computation or invoke external grid/cloud emulators
via plugins
Post computation - Data production registration
Output data transfer - Capture the data transfers that might be required for
the output data generated by the emulated flow.
Task complete - This is the final stage, execution flow results are sent back
to the workflow engine.
We can use pipeline/chain implementation to run this flow steps. Some states
are not valid for some flows, such states can be interrupted by user. eg:
Output data transfer state is not valid when there is no any output from
emulated computation.
Inside the Computation state user can define failure rate.
As for the ease of recording, I will be updating this in the JIRA too.
regards,
Shameera.
> [GSoC] Framework to run/monitor workflows locally with simulated/dummy BE
> -------------------------------------------------------------------------
>
> Key: AIRAVATA-346
> URL: https://issues.apache.org/jira/browse/AIRAVATA-346
> Project: Airavata
> Issue Type: New Feature
> Reporter: Saminda Wijeratne
> Assignee: Saminda Wijeratne
> Labels: gsoc2012, gsoc2013, mentor
>
> It is common to need to test something you've created on a test environment
> rather than on a real setup. This is because,
> 1. The tasks are trivial and it'll be faster to run locally to verify correct
> flow/execution.
> 2. It can cost to run them on a real setup at development stage.
> 3. Proper access to the real setup is not available
> 4. Unit testing on a real setup is unfeasible
> Thus it would be easy for the user if there is a way to setup a simulated
> environment without much difficulty may be perhaps with some dummy data from
> the user itself. Effectively it will be behaving exactly as a GFac
> server/message box/registry etc but with hard wired data. As an example
> ideally it would be like the Azure simulator I guess.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
