Testing processes for efficiency

Abstract

Two notions for comparing the efficiencies of equivalent concurrent systems have been developed and axiomatized in [1] and [2]. Recently Natarajan and Cleaveland have defined a notion of testing [6] which incorporates these ideas as an extension of the testing methodology ([3],[4]). Their extension bounds the run of a test in an effort to compare processes for their relative efficiencies, measured in terms of the amount of internal activity in a process.

In this paper we explore the feasibility of obtaining other efficiency based preorders in the setting of the testing methodology. We generalize the testing methodology and cast [3] and [6] in the new framework. Further, we provide a variation that is equivalent to that of [6].

Another alternative that we explore is what we call “testing for efficiency”. Rather than bound the test runs externally, testing for internal activity in a process is woven into the test itself. This turns out to be a more powerful and flexible method of testing processes for a variety of properties.

The main result of this paper is that the testing-efficiency preorders obtained from all the variations previously considered are coarser than that obtained by “testing for efficiency”. We prove appropriate alternate characterizations that are independent of any testing formalism. We also show that the existing methodologies (including [3]) or their equivalent variations can be simulated in the “testing for efficiency” framework.

Supported by project grant SR/OY/M-08/92 of the Department of Science and Technology, Govt. of India.