A sequential probability ratio test (SPRT) is discussed, for comparison of two systems, one "basic" (b), and the other "new" (n), with exponentially distributed times between failures (TBF). The hypothesis that the mean TBFn/MTBFb ges 1 is checked, versus one that it is <1. The paper deals with tests with a low Average Sample Number (ASN), having the advantage of economy in time requirement, and cost; and it is shown that the points of possible solutions in them are sparse. Criteria are proposed for assessment of the test quality, with a view to optimization of its parameters. We present a search algorithm for the truncation apex (TA), with dependences for the search domain, and for the position of the oblique test boundaries, serving jointly as the basis for our development of the test planning algorithm.