Novel clock synchronization algorithm of parametric difference for parallel and distributed simulations

Abstract

Clock synchronization is crucial in distributed simulation applications (DSAs). Traditional synchronization methods are mainly based on computers’ physical clocks and have network delays. Consequently, the clock jumps easily, and the stability of the algorithm is less than excellent. Furthermore, a new class of DSAs with flexible and dynamically assembled simulation components to optimize the performance of clock clusters is emerging. To remedy the problems, we propose a novel logic parametric difference clock (PDC), which is recursively constructed by an alterable parametric difference frequency (PDF) and its counter. We also develop a simple and efficient synchronization algorithm of parametric difference (SAPD), in which the indeterminate network delay is considered in the PDF, and the PDC counter is used as the synchronous symbol. Compared with existing typical algorithms, the SAPD has the predominant advantages of convergence, stability, and precision for DSAs in local area networks.

Introduction

We consider the clock synchronization issues of parallel and distributed simulation systems with dynamically assembled components in an Ethernet network. In recent years, large-scale distributed applications (e.g., grid computing, cloud computing, pervasive computing, and Internet of Things) have become a major trend in distributed systems [1], [2], [3], [4], [5] mainly because of the enormous development of information technologies, power technologies, sensor networks, and systems on a chip. Nowadays, more distributed application systems are increasingly characterized by novel work mechanisms, brand-new physical architecture, and dynamic topology of nodes [6]. Fig. 1 shows this kind of heterogeneous system and its novel applications. The next generation’s distributed or simulation systems are predicted to exceed traditional ones in local area networks (LANs) and wide area networks (WANs) because of the former’s predicted distribution, generality, gridding, and service-oriented traits. Especially in simulation fields, service-oriented composable distributed simulation applications (DSAs) are becoming a popular trend [7].

In general, some of the major features of emerging distributed application systems or DSAs are as follows [6]:

• Generality. The novel system is more suitable for constructing a common distributed platform such as Grid and Cloud environment to meet all kinds of requirements than the systems in other application fields. That is, the system, supported by first-class hardware and software, is capable of giving a package solution to many applications.

• Dynamic assemblage. The topology of system nodes, which is dynamically made up of unknown distributed nodes, is transformable, unpredictable, and therefore stochastic.

• Redundancy. Various redundant measures are utilized to strengthen the robustness of the system because of the unstable nodes in new distributed systems.

• Staged synchronization. The dynamic changes in topology indicate that many nodes are often synchronized at the beginning and end of some tasks.

Clock synchronization is one of the most fundamental issues of each distributed application system or DSA, an age-old problem [8], [9] that many researchers have studied. To enable clock consistency in such simulation environments, a scalable and common clock synchronization service is required.

In prior studies, many effective methods, such as network time protocol (NTP) [10] in traditional Ethernet, precision time protocol (PTP) [11] in industry control, and Global Positioning System (GPS)-based synchronization [12], are presented to solve this problem. However, these methods have many flaws, such as network delays, missing packages, and high cost. For instance, WSNs have many protocols and algorithms, such as reference broadcast synchronization (RBS) [13], flooding time synchronization protocol (FTSP) [14], and timing synch protocol for sensor networks (TPSNs) [15]. Considering the specific requirements of low battery, durative stability, high security, and narrow bandwidth, many scholars have proposed extended methods to improve the efficiency, robustness, and security of their algorithms [16], [17], [18], [19], [20], [21], [22], [23]. However, these algorithms [24] are based on the characteristics of WSN, and the synchronization mechanisms of these algorithms are not well applied to DSAs in LANs because several indefinite factors exist for cluster synchronization in traditional Ethernet and multitask operation systems, such as all kinds of delay in network transport, operation system schedules, packing and unpacking, and function calls [25].

With the popularity of this new DSA class, time synchronization exhibits some new problems (e.g., decreased stability, lower accuracy, and high spending of synchronization) with regard to novel distributed application systems. Moreover, traditional synchronization protocols and algorithms have difficulty keeping clocks in phase with the general application platform of distributed simulation, especially when the entry and exit of simulation nodes are frequent. To implement steady time synchronicity in such an extended distributed simulation environment, a simple, flexible, and efficient algorithm is required.

Considering the features of novel distributed application systems and investigating the advantages and disadvantages of classical synchronization algorithms for traditional networks and WSNs, we propose a novel logic clock model and synchronous scheme that does not modify the physical clock values of local computers. The goal of our algorithm is to explore and resolve the clock synchronization problem of new DSAs (e.g., grid-based and service-oriented DSAs) in the Internet from a new perspective.

The main contributions of this paper are as follows:

• We introduce a new logic parametric difference clock (PDC) model of heterogeneous clock clusters in our self-designed distributed simulation middleware (run-time infrastructure [RTI] based on service-oriented architecture, which is similar to high-level architecture/RTI).

• We mainly propose a novel clock synchronization algorithm of parametric difference (SAPD) that uses PDC in distributed simulation systems but can be used in real distributed systems in LANs. The SAPD sufficiently considers the dynamic changes in systems, the essence of time synchronization, and the staged synchronization requirements of fluctuant nodes. Moreover, by adopting the broadcasting and parametric difference of RBS [13], the interactive synchronicity of the reachback firefly algorithm (RFA) [26], and the multi-broadcast estimation of cooperative time synchronization (CTS) [27], the SAPD provides higher stability within microsecond accuracy and a brand-new method for time synchronization by calibrating the parametric difference frequency (PDF) stringently according to the frequency wave.

• We analyze the PDC model and the SAPD theoretically and experimentally and validate their feasibility and efficiency through simulation measurements.

The remainder of the paper is organized as follows. In Section 2, we expand the related research on time synchronization in distributed systems, including WSNs and grid systems. In Section 3, we propose a new clock model, PDC, and a novel time-calibrated algorithm, SAPD, which is completely different from existing algorithms. In Section 4, we validate the performance of the proposed algorithm by analyzing the results of the simulation experiments. In Section 5, we provide the conclusion and the recommendations for future work.