Next: JXTA
Up: Peer to Peer Computing
Previous: SETI@home
Contents
Parabon Computation, Inc. [48] announced a commercial
distributed application called Frontier at the Supercomputing 2000
conference. The Frontier is available as a service over the Web or as service
software. It claims to deliver scalable computing capacity via the
Internet to a variety of industries.
The Client Application, the Pioneer Compute Engine, and the
Frontier Server are three components of the Frontier platform. The
client application runs on a single computer by an individual or organization
wishing to utilize the Frontier platform by communicating with the Frontier
server. The Pioneer Compute Engine is a desktop application that utilizes the
spare computational power of an Internet-connected machine to process small
units of computational work called tasks during idle time. The Frontier Server
is the central hub of the Frontier platform, which communicates with both the
client application and multiple Pioneer compute engines. It coordinates the
scheduling and distribution of tasks.
Frontier gets a ``job''--defined by a set of
elements and a set of tasks to perform computational work--from user.
Then like work units in SETI@home, a job is divided into an arbitrary set of
individual tasks that is executed independently on a single node. Each task is
defined by a set of elements it contains a list of parameters, and an entry
point in the form of a JavaTM class name.
The inherent security features of JVM technology was the main reason to
choose Java over other programming languages as programming language of
Frontier. The JVM provides a ``sandbox'' inside which an engine can securely
process tasks on a provider's computer. Valid Java bytecode has to be sent to
engines and used to run tasks within the JVM.
In this project sustaining a large network of ``volunteer'' machines is a
problem. Not many consumers are willing to donate computing cycles for
purely
commercial projects. Parabon system is good for task parallel applications
but arguably
is not peer to peer computing in sense tasks cannot communicate. Parabon
system is less appropriate for the more tightly-coupled SPMD programming
we are interested in.
Next: JXTA
Up: Peer to Peer Computing
Previous: SETI@home
Contents
Bryan Carpenter
2004-06-09