5 Implementation
A Distributed Maple Session
The architecture of a distributed Maple session is sketched in
Figure 5. Each node connected to the session consists of two
components:
- Scheduler
- A scheduler program manages the node
interaction and schedules tasks among nodes. This program is implemented by a
Java class library with main class dist.Scheduler and is independent of
Maple. The initial scheduler process reads all application-specific
information from the configuration file
dist.systems and may then start instances of the scheduler on other
machines with which it communicates via sockets.
- Maple Interface
- The package dist.maple running on each Maple
kernel implements the interface between Maple and the scheduler. Communication
between both components is based on pipes (named pipes for the Maple kernel
connected to the user interface and standard input/output streams for
the backend kernels).
During a session, additional socket connections between remote scheduler
instances are dynamically established on demand.
Maintainer: Wolfgang Schreiner
Last Modification: July 6, 2001