airavata-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gourav Rattihalli <gratt...@binghamton.edu>
Subject [GSoC Proposal] - Integrating Job and Cloud Health Information of Apache Aurora with Apache Airavata
Date Mon, 21 Mar 2016 14:22:31 GMT
Hi Dev Team,

Please review the following GSoC proposal that I plan to submit:

*Title*: Integrating Job and Cloud Health Information of Apache Aurora with
Apache Airavata

*Abstract*:

This project will incorporate Apache Aurora to enable Airavata to launch
jobs on large cloud environments, and collect the related information on
the health of each job and the cloud resources. The project will also
analyze the current micro-services architecture of Airavata and develop
code for an updated architecture for modules such as Logging. As as result,
another outcome of this project would be development of a module that will
collect all the logging information from the various execution points in an
Airavata job's lifecycle and provide search and mining capability.


*Introduction*:

Apache Aurora is a service scheduler, that runs on top of Apache Mesos.
This combination enables the use of long running services that take
advantage of Apache Mesos scalability, fault-tolerance and resource
isolation. Apache Mesos is a cluster manager, which provides information
about the state of the cluster. Aurora uses that knowledge to make
scheduling decisions. For example, when a machine experiences failure
Aurora automatically reschedules those previously-running services onto a
healthy machine in order to keep them running. Each job is tracked by
Aurora to be in one of the following states: pending, assigned, starting,
running, and finished.

Apache Aurora requires a configuration file ”.aurora” to launch jobs.
Following is an example of Aurora configuration file:

import os
hello_world_process = Process(name = 'hello_world', cmdline = 'echo hello
world')

hello_world_task = Task(
 resources = Resources(cpu = 0.1, ram = 16 * MB, disk = 16 * MB),
 processes = [hello_world_process])

hello_world_job = Job(
 cluster = 'cluster1',
 role = os.getenv('USER'),
 task = hello_world_task)

jobs = [hello_world_job]

To launch the job with the above configuration we use:

aurora job create cluster1/$USER/test/hello_world hello_world.aurora

This project will develop modules in Airavata to automatically generate the
Aurora configuration file to launch a job on an Aurora-managed cluster in a
cloud environment. The Aurora user interface, as shown in the web portal
displayed above, provides detailed information on the job status, job name,
start and finish times, location of the logs, and resource usage. This
project will use add a module to Apache Aurora to pull this detailed
information using the the Aurora HTTP API.

*Goals*:

   -

   This project will investigate how apache Aurora collects information of
   cluster environment for display on the Aurora web interface. We will study
   the Aurora HTTP API and retrieve all the information related to the target
   infrastructure and job health, and make it available to the Airavata job
   submission module.
   -

   We will process the retrieved information from Aurora and convert the
   information in a format that can be used by Airavata for further action.
   -

   We will use the appropriate design patterns to integrate the use of
   Aurora as one of the options for Big Data and Cloud resource frameworks
   with the Airavata framework
   -

   We will make the resource information from Aurora available for display
   on the Airavata dashboard.


Any comment and suggestions would be very helpful.

-Gourav Rattihalli

Mime
View raw message