The Connectivity Problem in Testing
With distributed processing, transactions are no longer processed
in a single system or environment, but instead are processed
across multiple platforms and multiple systems.
This distributed environment raises the level of complexity for application
programmers trying to test their transactions. Because of
this new complexity, to test a transaction a programmer needs
a complex testing environmentone that includes multiple
networks, multiple platforms, and multiple systems.
Without this complex level of connectivity, only parts of the
transaction can be tested, not the entire transaction. In addition,
the programmer must try to simulate the response from the
other environment to continue testing. However, setting up simulation
routines and test cases takes a great deal of time and makes programmers
Any solution to this testing problem has to
take two things into account.
First of all, each programmer requires a certain level of isolation to reliably
test application transactions. Second of all, at the same time the programmer
requires connectivity to other networks, terminals, systems, and so forth
for testing the remaining parts of the transaction.
One approach is to provide each programmer with a complete network,
terminals, systems, and so forth. This naive approach would be very
expensive, however, and would require a very large amount of resources,
as well as duplicate sets of resources. Another approach would be to isolate
the testing routines and provide server systems that do some of the common or
already tested functions. However, this approach puts a great deal of stress
on the testing organization and requires constant upkeep.
In our mind, neither of these approaches is satisfactory. What is needed is
an in-depth test facility where programmers can test their transactions using
enhanced trace, edit features, and source level trace features. At the same time,
these programmers should have nearly transparent access to other networks or
systems where certain sub-functions of the transaction being tested can be
performed. And all this should be provided without intensive use of resources.
CMSTPF and TPF/GI already
provide programmers with the powerful trace, edit and source
level trace features with a minimum use of resources. The
need for network connectivity is where CTFS steps in. With
CTFS and a TPF server machine connected to a network, CMSTPF
and TPF/GI programmers can perform the sub-functions (network
functions, and so forth) needed by their transactions. Complete
testing is possible without intensive resource use.