Scribd
Upload
80 views15 pages

Final Phase 3 Presentation PDF

This project aims to simulate computational offloading from an Android application to an AWS cloud server. The cloud server will use distributed computing to complete a compute-intensive task of searching a 1,000,000 element array in parallel. StarCluster will be used to automate and simplify management of the AWS EC2 cluster. Message Passing Interface (MPI) will handle communication between processes on different nodes to distribute the task. The Android app will connect to the cloud server via sockets to offload computation.

Uploaded by

Kumara S
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
80 views15 pages

Final Phase 3 Presentation PDF

This project aims to simulate computational offloading from an Android application to an AWS cloud server. The cloud server will use distributed computing to complete a compute-intensive task of searching a 1,000,000 element array in parallel. StarCluster will be used to automate and simplify management of the AWS EC2 cluster. Message Passing Interface (MPI) will handle communication between processes on different nodes to distribute the task. The Android app will connect to the cloud server via sockets to offload computation.

Uploaded by

Kumara S
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Distributed

Cloud Computing
with StarCluster
(DC2S)
CHRISTIAN ANGELES
JOHN ZAVALA
Project Description

► Our goal is to simulate computational offloading of a compute-intensive task on an android


application to a cloud server.
► Server will complete the task via distributed computation.
► Data will be an array of at least 1,000,000 elements conceptualizing a large data set--data that is
too large to store on an android device.
► Program on the cloud server will be performing an exhaustive search and comparison to simulate
the computational offloading.
► In order to create a compute-intensive environment, the algorithm is going to search and
compare the 1,000,000-element-sized array with itself. Forcing the algorithm to perform with a
time complexity of O(n^2). A slower time run will allow us to see a difference in computational
performance.
Project Environment

► Amazon Web Services(AWS) EC2


► Starcluster AMIs
► Android OS
► Linux OS
► StarCluster
► Message passing interface(MPI)
► mpi4py for Python
StarCluster

► StarCluster- Is a tool kit that manages computer clusters hosted on Amazon’s EC2
Cloud. It is designed to automate and simplify the process of building, configuring, and
managing clusters of virtual machines.
StarCluster

► Security Groups- The starcluster configures a security group for our cluster allowing us
to control network access to the cluster.
► Password-less SSH: The star cluster configures the cluster in order for the SSH to be
used from any node within the cluster without having to login using a password.
► Network File System (NFS)- able to attach Elastic Block Storage (EBS) volumes on the
cluster in order to obtain persistent storage.
► StarCluster can dynamically resize clusters in order to make it scalable.
► Combination of OpenMPI and Sun Grid Engine (SGE)
Message Passing Interface(MPI)

► MPI- Standardized message passing library interface that can be used in different
languages (ex. C, C++, python).
► MPI is good for high performance computing and parallelism because it handles the
passing of messages between different processes.
► Distributed computing- Takes a single task and distributes it among computers in order
to complete a task.
► Point to point communications- MPI uses point to point communications to
communicate from one process to another.
► Blocking/Non-blocking communication
► Race conditions
Message Passing Interface(MPI)

► Collective communication- There are two types: Broadcasting and


Scattering/Gathering.
► Broadcasting: One process broadcasts the same information to every process.
Message Passing Interface(MPI)

► Collective communication- There are two types: Broadcasting and


Scattering/Gathering.
► Scattering/ Gathering: Are used in order to distribute and gather data.
Multithreading
➤ Multithreading is an execution model that
allows multiple threads to be created within a
process such that the threads execute
independently but share the same process
resources.
➤ Every thread contains a unique register set
and local variables. (Local variables are stored in
the stack).
➤ All the threads within the process share the
same global variables and code.( The global
variables are stored in the heap).
➤Main reason we are using multithreading
rather than multiprocessing is because of less
overhead. It’s a more efficient way of performing
the task.
Cloud Server

► Written in python for easy implementation


► MPI support with mpi4py
► Socket programming
► Multiprocessing/Multithreading libraries
► Main process
► Listens for client connection
► Creates a thread for each client connection
► Launches separate process for distributed computing
► MPI process
► Processes created for each node
► Divides array into segments
► Creates multiple threads for each segment
Android Application

► Application development in Android


Studio.
► Designed a simple user interface
► Single button to start
► Contains implementation of socket
programming in order to establish a
connection to the cloud
► Creates a separate thread for connection
to the cloud server
► AsyncTask
Demonstration
Lessons learned
► AWS Elastic Cloud Computing (EC2)
► Creating instances
► Configuring network security group
► Allows android application to
connect
► StarCluster
► Parallel Computing
► MPI(Message Passing Interface)
► SGE( Sun Grid Engine)
► Multiprocessing/Multithreading
► Android
► Python
References

► StarCluster
► http://star.mit.edu/cluster
► Distributed computing with MPI
► https://www.codingame.com/playgrounds/349
► Open MPI
► https://www.open-mpi.org
► mpi4py
► https://mpi4py.readthedocs.io
Thank you!

Questions?

You might also like