features for the dataflow-based integration framework include:
1. Ease of use
2. software reuse - 
3. software modularity
4. scalability

Process-based distributed integration environments (AVS, SGI
Explorer, IBM Data Explore, Khorus etc), although originally designed
for workstation operation, the framework provided by these systems
offer a means by which distributed systems may be connected
by the worksations or high performance computing systems.

DVEs are characterized by a user-built framework of interconnected
modules representing the dataflow for a given application. Such systems
support rapid prototyping, interactive application steering, and device independence.
from the application writer's perspective, this approach relieves the 
programmer from having detailed understanding of graphics and offers
significant software savings through the resue of modules. Because
the interfaces between modules are well-defined, modules developed
at different computer sites/platforms may be freely shared.

The dataflow network, constructed at run time supports distributed computing in a
heterogeneous LAN or WAN network environment of high performance systems
and graphics workstations. Modules themselves perform the underlying work of data
input, feature extraction, data mapping, rendering, and output. DVE's hide the
inherent complexity of different internal numerical representation, bye ordering,
vector/scalar/parallel visualization/computation algorithm dependencies, or
serial/parallel I/O constructs in a modular HPC
heterogeneous framework. Overall, DVE's give the user control over the
optimization of heterogeneous resources and network capabilities.

Users may interactively steer simulations through adjustment of an application's 
critical parameters. In a similar manner, modification and adjustment of the
process itself may be performed in an interactive manner. The same
framework which facilitates the easy creation of complex application taskes
also provides for its quick modificaiton to include new and different data mapping
and feature extraction functions.

The modular approach of DVE's lend themselves well to software sharing, module
reuse, extensibility and flexibility. Software sharing means that modules
created for one high performance architecture can be readily to ported
to another similar architecture. The reusability aspect of DVE's means the HPC 
parallel code and algorithm are modularly developed and compiled once, then reused
indefinitely in a plug-and-play mode in a variety of diverse applications. The
extensibility of DVE framework is such that users may write their
own modules in their favorite HPCD langauge/program binding.