ALTIMETRIK

Grid Computing on
Mobile Devices

A Point of View

Sriraman K R, Lead Architect,
Research & Innovation, Altimetrik

This paper explores the possibilities of using mobile devices like
smartphones and tablets in the area of grid computing / distributed
computing, talks about progress that has already been made in this
direction and outlines techniques to be adopted for a successful
implementation.
ALTIMETRIK
Altimetrik is a rapidly growing
organization with a vision to
enable business innovation
and heighten user
experience by leveraging
next generation technologies
and design led engineering
principles. This is achieved by
incorporating user behavioral
patterns into its solutions using
accelerated design and
deployment methods






















Grid computing or distributed computing refers to a technique in
which a central server breaks a large computational problem
into a set of independent workloads and sends it to a number of
connected autonomous computing nodes having their own
local memory, storage and CPU. These nodes then operate on
the workload using their own computing resources, gather the
computational results, send it back to the parent server and
download a new set of workload for processing.

Computing resources required to solve huge computational
problems are very expensive and complex. A cost effective
alternative is to distribute these workloads to a large number of
personal computers, whose owners donate the idle CPU time of
their computers. Programs like SETI@home have popularized
these techniques and have given rise to a new concept called
Volunteer computing.

This paper explores the idea of applying the above technique to
mobile devices and explores the nuances of using them as end-
computational nodes in a grid computing system.
3.1 Factors and Opportunities

The advent of mobile phones has created exciting opportunities
and possibilities in the potential development of grid / distributed
computing techniques, especially in the current era of powerful
smartphones and tablet devices. With over 10 million
smartphones out there today and further expected to grow,
smartphones pose a potential platform for grid or distributed
computing for the following reasons:

CPU Most of the smartphones (and tablet devices likewise)
today have at least 1 GHz of processor speed, upwards of 256
MB of RAM and these specifications are ever increasing.
Connectivity Wider geographical reach of mobile networks
and advancements in technologies like GPRS, 3G, 4G and Wi-Fi.
Storage Significant increase in the storage capabilities of
running into a few gigabytes.
Sensors Attachment of sensors like GPS, accelerometer,
compasses, camera etc. These sensors provide a context for the
distributed application.























However there exist two other key reasons that make
smartphones a compelling platform for grid or distributed
computing:

1. Almost always on
2. Most of the time, only a fraction of the computing power of
smartphones is ever used, thus providing enough CPU resource
to run distributed applications.
3.2 Challenges

There are a few important challenges that need to be
addressed before this can be successfully adopted, of which
four primary challenges are listed below:

Power - There is only a limited amount of power available on a
mobile device at any given time. Heavy computational tasks
will drain the battery almost rapidly.
Platform - Distributed applications targeted at mobile
platforms have to run on heterogeneous systems using a
variety of hardware, OS and libraries. Popular platforms like
Android are heavily fragmented further adding a level of
complexity.
Network - A mobile device is often intermittently connected to
a network and connects to a variety of networks including
cellular, data, Wi-Fi etc.
Social - Mobile device owners may not be compelled to have
their devices participate as their primary concern would be
battery drainage and connectivity charges that they may
incur, although contributing towards the bigger goal of solving
a large problem may be encouraging.

3.3 Approaches

Lets explore some of the possible approaches that can be
adopted to solve some of these challenges.

Distributed applications targeted at mobile platforms need to
be specifically optimized for that platform. A cubic millimeter
of current battery technology has enough energy to perform
about 1 billion 32-bit computations, or send and receive 10
million bits of data. Thus distributed applications need to follow
efficient algorithms and programming techniques that
optimize battery consumption.






















However, ongoing improvements in energy densities in
batteries could solve this situation in the future. Alternatively,
mobile device owners eager to participate in a grid computing
exercise may choose to plug-in their devices to external power
outlets.
To target heterogeneous mobile platforms and the different
network topologies t h at the mobile devices typically connect
to, distributed applications needs to be sufficiently abstracted
from the underlying platform. Implementations can be on cross-
platform technologies like Java with an abstracted Java virtual
machine. Microsofts .NET framework is also available for both
iOS and Android platforms (through Novels Mono
implementations). However, this will still pose a challenge to
implementing platform specific algorithmic efficiencies and
different versions for each platform may have to be written.

An effective reward system needs to be in place to compel the
mobile device owners to participate in a grid computing
project and donate their devices computing resources. Project
owners of significant scientific or global importance can tie up
with mobile phone carriers and service providers for bundling
these distributed applications along with the devices. The
service providers can then track the usage and reward the
users directly with discount offers, cash back or similar
programs. Special rewards can be given to the owners of
devices that actually contributed directly to any results /
milestones achieved in solving a computational problem.

3.4 Current Implementations

There are various reference architectures and technical
frameworks already in place for efficiently distributing work
across a set of mobile devices and for using them as
computational nodes, many of the projects being outputs from
academic research initiatives.

Ibis, a research project from VU University, NL, is an open source
Java based high performance distributed computing platform.
It has been successfully ported to Android smartphone
platform.























OGSI.NET is a grid computing project from University of Virginia. It
is an implementation of the OGSI (Open Grid Service
Infrastructure) specification on Microsofts .NET platform. Mobile
OGSI.NET extends this implementation of grid computing to
mobile devices and tries to address a few challenges like
resource limitations and intermittent network connectivity.

AVRF Android Volunteered Resource Framework enables
distributed computing using volunteered mobile devices. The
goal of this project was to create a framework to enable
Android smartphones to act as volunteer workers for distributed
computing tasks.

The Berkeley Open Infrastructure for Network Computing
(BOINC) is an open source middleware system for volunteer and
grid computing which uses idle time on personal computers to
run computational problems. BOINC Mobile, aimed at BOINC
developers attempts to the BOINC platform and its projects
(SETI@Home, etc.) to ARM platform based mobile devices.
BOINCOID is a project that attempts to port BOINC to Android
devices.

One of the most popular projects is the Misco, a distributed
computing framework designed for mobile devices, which
provides a platform for developing distributed applications
targeted at mobile devices. It first began as an intern project in
the summer of 2008 at Nokia Research Center in Palo Alto. It is
implemented completely in Python and is based on the popular
MapReduce programming interface. It is still in its early stages
and is available for free public download.

The ideas presented in this paper are possibly an exploration of
some of the ideas of a futuristic concept called The Ambient
Cloud.

An Ambient Cloud is a massively distributed collection of
computer resources constructed from potentially any device,
where these resources would be found. It will not be controlled
or governed by a standards body nor will it be a traditional
billion dollar datacenter project. It would be held together by
an informal, tacit and negotiated relationship and would be
irregular in connectivity, policy, cost, technology and capability.
The Ambient Cloud would be bound together by applications
and not platforms, APIs or services.





















The sheer number of mobile phones out there will provide large
scale computational power and if augmented with the
existing personal computers participating in programs like
Volunteer computing, will further increase it exponentially.
This approach has the potential to deliver computational
power way beyond the fastest of the super computers that
exist today at a fraction of the cost, with practically no
infrastructure investments required from the project owner.
There is also the added advantage of savings on energy
consumption and smaller ecological impact.

The sensors that come with the mobile devices further give a
context to the distributed applications and opens up a new set
of possibilities. Imagine, a mobiles GPS sensor can be used by
a distributed weather pattern analysis program for accuracy
over a micro-geographical region or used for analyzing traffic
patterns over a geographical region. The microphone of the
mobile can be used for analyzing, say, noise levels over a
region, at different times of a day. Projects like Quake-Catcher
Network use accelerometer equipped laptops to detect
seismic activity and earthquakes, which can be extended to
mobile devices as most smartphones today are equipped with
accelerometers.

Its quite a long way off before we have mobile-powered grid
computing infrastructures but the capability exists and its just
time before more applications emerge. This could potentially
lead to the beginning of the Ambient Cloud.





















References

Arnold, E. (2011). AVRF: A Framework to Enable Distributed
Computing Using Volunteered Mobile Resources. Retrieved from
University of Puget Sound:
http://soundideas.pugetsound.edu/summer_research/127/
BOINC. (n.d.). Retrieved from http://boinc.berkeley.edu/
BOINC Mobile. (n.d.). Retrieved from Google Code:
http://code.google.com/p/
boincmobile/ BOINCOID. (n.d.). Retrieved from SourceForge:
http://sourceforge.net/projects/ boincoid/
Chu, D. C., & Humphrey, M. (2005). Mobile OGSI.NET: Grid
Computing on Mobile Devices. Retrieved from University of
Virginia: http://www.cs.virginia.edu/~gsw2c/ ogsi.net.html
Hoff, T. (2009, December 16). Building Super Scalable Systems.
Retrieved from High Scalability:
http://highscalability.com/blog/2009/12/16/building-super-
scalable- systems-blade-runner-meets-autonomic.html
Hoff, T. (2010, August 18). Misco: A MapReduce Framework for
Mobile Systems - Start of the Ambient Cloud? Retrieved from
High Scalability:
http://highscalability.com/blog/2010/8/18/misco-a-
mapreduce-framework-for-mobile-systems-start-of-the.html
misco - A Mobile MapReduce Framework. (n.d.). Retrieved from
University of California: http://www.cs.ucr.edu/~jdou/misco/
Palmer, N., Kemp, R., Bal, H., & Kielmann, T. (n.d.). IBIS. Retrieved
from vrije Universiteit: http://www.cs.vu.nl/ibis/papers.html
Quake Catcher Network. (n.d.). Retrieved from Stanford
University: http://qcn. stanford.edu/ Wikipedia. (n.d.).
Distributed Computing. Retrieved from Wikipedia:
http://en.wikipedia.org/wiki/Distributed_computing
Wikipedia. (n.d.). Grid Computing. Retrieved from Wikipedia:
http://en.wikipedia.org/ wiki/Grid_computing
Wikipedia. (n.d.). Volunteer Computing. Retrieved from
Wikipedia: http://en.wikipedia. org/wiki/Volunteer_computing
















About the Author

Sriraman K R, Lead Architect, Research & Innovation at
Altimetrik leads the Research & Innovation group which is
responsible for enterprise business focused blue sky technology
innovation, advanced technology solution strategy and
incubation leading to IP creation, Patents and other non-linear
value multipliers.

His special interests include disruptive technologies, software
and enterprise architectures, service oriented architectures and
start-ups. He is an early entrant in technology areas like Web 2.0,
cloud computing and has authored papers on these subjects.