|
These years, the research in the realm of distributed shared
memory (DSM) has been a hot research area. The benefits of DSM
over that of message passing mechanisms are widely
acknowledged. Instead of involving in the research about DSM
meanwhile, a varied meachanism, distributed shared variable
(DSV), will be applied because the development of DSM always
requires extra modification for the kernel of operating system.
The major differences between DSV and DSM paradigm can be
distinguished in the following. DSV paradigm only supports the
data sharing in a distributed computing environment. However,
DSM paradigm can support data and code sharing. Nevertheless,
the source codes of the existing widely accepted commercial
operating systems are not affordable easily. So, pursuing a
method with no kernel modification and with capability of
achieving data sharing in heterogeneous computing environment
is our final reasonable choice. DSV is not only an efficient
programming interface, but also a friendly high-level
application programming interface. While the scale of software
will grow large, the DSV paradigm should be able to cure the
original crisis incurred by software complexity. PVM is a user-
level software system which integrates lots of existing famous
computing resources on the network into one large virtual
entity, however, it lacks an efficient programming paradigm
such as shared variable or even shared memory. After taking the
above into account, we choose this platform, PVM system, to
develop the prototype of DSV system as quickly as possible.
|