[jira] [Commented] (SINGA-11) Start SINGA on Apache Mesos

["Anh Dinh (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SINGA-11?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14962175#comment-14962175 ] 

Anh Dinh commented on SINGA-11:
-------------------------------

pull request created:

https://github.com/apache/incubator-singa/pull/85

this one depends on SINGA-89 whose pull request has been created. 



> Start SINGA on Apache Mesos
> ---------------------------
>
>                 Key: SINGA-11
>                 URL: https://issues.apache.org/jira/browse/SINGA-11
>             Project: Singa
>          Issue Type: New Feature
>            Reporter: wangwei
>            Assignee: Anh Dinh
>
> Apache Mesos is a fine-grained cluster management framework which enables resource sharing in the same cluster. Mesos abstracts out the physical configurations of cluster nodes, and presents resources to the users in the form of "offers". SINGA uses Mesos for two purposes:
> # To acquire necessary resources for training the model.
> # To launch and monitor progress of the training task.
> To these ends, we implement a {{SINGA Scheduler}} which interacts with Mesos master. The scheduler is called when the user wants to start a new SINGA job, and it performs the following steps:
> # Read the job configuration file to determine necessary resources in terms of CPUs, memory and storage.
> # Wait for resource offers from the Mesos master.
> # Determine if the offers meet the requirement of resources.
> # Prepare the task to launch at each slave:
> #* Deliver the job configuration file to the slave node.
> #* Specify the command to run on the slave:
> {code}
> singa -conf ./job.conf
> {code}
> #* Launch and monitor the progress
> For step 3, we currently implement a simple scheme: the number of CPUs offered by each Mesos slave exceed the total number of SINGA worker and SINGA server per process. In other words, each selected slave must be able to run the entire worker group or server group.
> For step 4, we currently relies on HDFS to deliver the configuration file to each slave. Particularly, we write the file to a known directory (different for each job) on HDFS and ask the
> slave to use its Fetcher utility to download the file before executing the task.
> The development and testing environment for this ticket are created from [SINGA-89|https://issues.apache.org/jira/browse/SINGA-89]
> We will create a {{README.md}} file explaining the steps.
> h5. Important
> We assume that SINGA, Mesos and Hadoop are running at every node. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)