A concept of the ground-based optical astronomical observations efficiency is considered in this paper. We believe that a telescope efficiency can be increased by properly allocating observation tasks with respect to the current environment state and probability to obtain the data with required properties under the current conditions. An online observations scheduling is assumed to be essential part for raising the efficiency. The short-term online scheduling is treated as the discrete optimisation problems which are stated using several abstraction levels. The optimisation problems are solved using a parallel depth-bounded discrepancy search (PDDS) algorithm [13]. Some aspects of the algorithm performance are discussed. The presented algorithm is a core of open-source chelyabinsk C++ library which is supposed to be used at 2.5 m telescope of Sternberg Astronomical Institude of Lomonosov Moscow State University.