عنوان مقاله
معماریهای ارتباطی برای سیستم کنترل کامپیوتری توزیع شده
فهرست مطالب
مقدمه
معماری پایه DCCS
نیازمندیها و اثرات وارده بر معماری سیستم
پروتکل های استاندارد: کاربرد معمولی
مفاهیم چند بخشی برایDCCS
نتیجه گیری
بخشی از مقاله
معماری پایه DCCS
سیستم های کنترل کامپیوتری صنعتی مدرن به شکل سیستم های توزیع شده طراحی شده اند (شکل 1). سیستم های در نظر گرفته شده در اینجا از تقریباً 15- 10 کامپیوتر تابعی (عملکردی) تشکیل می شوند که مدول های تابعی و کارکردی توصیف شده در قسمت ذیل برای آنها می توانند تصادفاً باهم ارتباط داشته باشند.
کلمات کلیدی:
COMMUNICATION ARCHITECTURES FOR DISTRIBUTED COMPUTER CONTROL SYSTEMS W. Dieterle*, H.-D. Kochs* and E. Dittmar** *Department of Computer Science, University of Duisburg, Lotharstr. 1, 47048 Duisburg, Germany **ABB Netzleittechnik Gmbh, Network Control and Protection, Wallstadter Str. 53-59, 68259 Ladenburg, Germany (Received October 1994; in final form May 1995) Abstract: The use of distributed computer control systems (DCCS) demands high reliability. adequate real-time behaviour and increasinsJy econonxical systems. The last d t~nand rezluires the use of cheap standard ccmpcoents, wl~ncver possible. The following paper discusses the realizMJon of DCCS with respect to these constraints. Problems due to the c(mventional use of standardized c.¢wnmmfication prot(x~ls in distributed control systems in general, and highly-w.liable systems in particular, are shown. Multicast communication cxmcepts are ~ as solutions, using standardized protocols in a problem-specific way. The l~esented ccocepts fulfill the necessity of using standard components, as well as the specific &mands of DCCS. Keywords: Gontrol Systems. Local Area Networks. Computer Commnnieation Networks. Distributed Databases. Communication Protocols. Standard Protor~ls. 1. INTRODUCTION Hish ~mA,a~ are placed on distributed c~nputer control sys~us (DCCS). used in energy distribution. production or process eusineed~, whereby the costs aspect is more and mote dorninsmt~ Cost minlmization makes the use of cheap, standardized components and the design of simple, modular system concepts mandatory. In the following, specific architectllral feaD.lll~ for SySi~ti o ncnmnnlc4ttion, types of data storase and fault-tolerance in DCCS are derived from the systean requirements described. Existin~ standard communication protocols, e.g. TCP/IP. UDP/IP or ISO/OSI. are not intended to support these features; however, the lack of appropriate standard protocols in the UNIX mrcironm~mt requires the use of the existin~ ones. Problems with the conventional use of standard _c~nmunicstion protocols are shown, and two multicast concepts are presonted and evaluated. They are based on standardized communication pretom~, but use them in a problm-speciec manner. The multicast concepts are very simple (in comparison with existin~ solutions) and have a very low message OVerPAid, close to the minimal message cost which is deterrnin,~l by simplified border conditions. Experimental results show that the timine characteristics of the first solution (ring multicast) are acceptable for small and medium-sized DCCS. The second solution (datagram multicast) is suitable for large DCCS and systems with specific demands for d_ata tr~n~er time and throughput. 2. BASIC ARCHH'EUTL~ OF DCCS Modem industrial computer control systems are designed as distributed systems (Fig. 1). The systems consklered here consist of approx. 10-15 functional computers, to which the functional modules described below can be randomly associated. Functional c.~n~ters are connected via Local Area Networks (LAN), typically Ethernet. To meet the high demands for reliability, computers with important functiom are redundantly s~. The func~onal scope of such systems incorporates data acquisition, b~s~_'_~ wocessing of process dam (SCADA). a,a process visuatizatim (MMI), as well as additional funcfimal modules (complex secondary functions) depmdin~ on the con-
