SCALABLE DATA PROCESSING IN AUV ONBOARD DISTRIBUTED COMPUTING SYSTEM

Abstract

Development of distributed computing systems (DCS) takes an important place in modern scientific and technical literature. Generally, only one of the most significant features of DCS is discussed in the papers, for example, performance, reliability, fault tolerance, energy efficiency and scalability. In this paper authors attempt to overall consider the problem of DCS design, based on the example of a multi-channel onboard DCS for data processing places in autonomous underwater vehicle (AUV). The aim of this paper is to formulate a unified concept of a multi-channel onboard DCS for real-time data processing. As a result, the architecture and principles of operation of a multi- channel onboard DCS are proposed, based on well-known approaches to fault tolerance and energy efficiency, taking into account the features of scalable systems. The proposed solutions can be viewed as an advancement of traditional approaches to scalable systems development. Fault tolerance is achieved by using test-based diagnostic tools. In order to reduce the complexity of these tools, redundancy is preliminarily added into each software module (SM) of the system. Then tests for the redundancies are built. It is shown that this test detects failures in the addressing of exchanges between SM blocks that implement the data processing. Based on the results of the analysis of the diagnostic tool reaction to the test, a failed software module is detected. Then failed module stops its work, and a new SM that implements the same algorithm is started execute instead. Energy efficiency proposals are suitable for the case of the presence of redundant processors in the system which could support multi-core technology. These processors could be involved in the execution process of the system SMs with a simultaneous decrease in the clock frequency and supply voltage. Since the power consumption in the DCS significantly depends on frequency and supply voltage, it decreases along with this parameter values. An optimal greedy algorithm is used to solve described problem, which assumes sequential involving of additional processors into the system. It is important that energy efficiency proposal of the DCS provides the latter additional fault tolerance capabilities. The practical importance of the proposed concept consists in the possibility of using not only in AUVs application. It also could be used in other cases of scalable multi-channel onboard DCS development with real-time data processing which have fault tolerance and energy efficiency requirements.