Three of the fastest growing technologies of all time—the wireless phone,
the handheld computer, and the Internet—are rapidly converging. Today 300
million people own a cell phone. Analysts predict that by 2003 as many
as one billion people will be carrying a device that is both a mobile phone
and a powerful computer wirelessly connected to the Net.
Vindigo's expertise is in building personal navigation tools, the killer
applications for these new devices. According to the global consultancy
Ovum, location based services will deliver $19.5 billion of revenue by
2006. Vindigo is at the forefront of this industry. Our award-winning software
and services can tell you where you are, what's going on in the world around
you, and how to find everything (and eventually everyone) in proximity
to you. (From http://www.vindigo.com/about/index.html)
Vindigo raised $6.1 million in May 2002 to expand its local area guides
Proposal: Evaluate Performance of Vindigo Server Architecture and Propose
The Vindigo server architecture consists of multiple tiers--database, application
server, and web server--with multiple machines at each tier. The system
is accessed in two modes: updates of synchronized clients (relatively large
and infrequent transactions) and online access for wireless users (relatively
small and frequent transactions). Vindigo expects significant growth in
online usage over time. They need to be able to measure the current capacity
of their system, tune the system where performance is poor, and come up
a way to estimate capacity as they add servers at different tiers. Vindigo's
goal is to support the most users at the least cost (both in hardware and
programmer/sysadmin time), and to be able to scale smoothly along with
growth in usage.
The proposed project has several parts:
1. Analyze Vindigo's usage data to come up with a practical model of
peak traffic on the system.
2. Set up load testing to measure current capacity.
3. Analyze bottlenecks at each tier to see where additional machines
can best be used at different capacity levels.
4. Develop a simple, peak capacity planning model for scaling to tens
of millions of users.
As the project progresses we will seek to identify software changes
which could improve performance. Changes may take many forms, including
DBMS tuning, commercial software replacement or upgrades, and data structure,
algorithm or protocol changes. The final report will contain a set
of suggested changes.
259 W. 30th Street, 4th Floor, New York, NY 10001
Company Web site
Bob Fitterman, Chief Technology Officer, firstname.lastname@example.org,
Jake Donham, Director of Technology, email@example.com,
Resources that will be made available to students
Access to any relevant data, code and information. Perhaps some hardware.