What is the Big Kahuna
• A Framework for performing massively
distributed computation using grid of stateless
transient nodes made up of mix of web
browsers, handhelds and thick clients which
make it a truly ubiquitous distributed
computing platform with no install or setup
Conceived - Summer of 2007
Originally a candidate for Product Prodigy 08
Paper submitted to WWW 09
Student initiated project at IIIT- Bangalore
A very common sight amongst Internet users.
So, why not get some useful work done whilst
the browsers are idle.
Problems with current technology
Thick App • Clients have to be downloaded and installed
Network • Ports have to be opened up on firewalls
Granularity • Not meant for small data sets
The Big Kahuna Solution
• Make the clients thin – remove the download
and install overhead
(which is usually untapped).
• Make the process seamless
• Zero installation /setup procedures
• Built on Open Source technologies
• Scales rapidly as more clients are added
• Extremely low startup overheads
• Problems modeled using the Map Reduce
• Platform and Browser independent
• Works on devices and phones. Truly
ubiquitous distributed computing platform.
The Big Kahuna is perfect …
• To Process large volumes of highly granular
data < 100Kb (ex: blogs )
• Solve a large subset of problems in IR and data
• Leverage idle time on the local network or the
• Deploy a distributed computing solution on
shared hosting servers
• If data is present on the client itself, eg
offline data on browsers.
HTML 5 proposes features for
Offline storage. Kahuna is
perfect to be the framework
for client side processing.
Market as a Market as a
A marketplace can The Kahuna Server
be setup, with will be available as a
scriptwriters and customized and fully
publishers as actors supported product
• Auto Tagging • Searching for
small blog 3x3 patterns in
articles using Chess Logs (log
Gears store and analysis)
What Big Kahuna isn’t ..
• It doesn’t guarantee speed up of operations,
primarily due to the network latency. It only
• Not targeted at datasets with large data unit
size (ex : multimedia data )
Barriers and Challenges
• The compute stubs are bound by the power of
• Bandwidth limitations hinder true
performance benefits of using the framework.
• Implement client side using other
technologies like flex/AIR, Silverlight etc