high-duty on-board computers
fault-tolerant on-board computers
programs of support of fault tolerance
V.A. Mihailov – Ph. D. (Eng.), General Director, JSC «NII «Argon». E-mail: firstname.lastname@example.org
V.A. Belov – Department Director, JSC «NII «Argon». E-mail: email@example.com
L.I. Dragomotschenko – Leading Engineer-Designer, JSC «NII «Argon»
A.A. Soloviov – Subdepartment Director, JSC «NII «Argon»
Already today practically operation with the digital video-information betrayed in real time on channels Fibre Channel on velocities боле 1 Gbit\s is required to all transportable automatic control systems. Video information usage in pattern recognition systems sharply boosts knocking abilities of tools of air defence. And optical data links allow to lower a visibility of operations control rooms of various level regarding deliberate interferences, that sharply reduces possibility of detection and an erasure of radio aids of transportable automatic control systems.
During development of units-servers following outcomes should be attained:
Critical lag ТТХ of domestic transportable complexes of an automatic control system and REB from foreign on productivity and reliability is lowered;
Creation of new domestic production engineering of development of the functional software for a new effective operating system of real time (OS РВ) for tasks of secondary information processing in air defence complexes is ensured.
Technical datas of the unit-server and composition of its hardware-software tools.
Functional composition of units:
Domestic 64 bit multinuclear microprocessors with MIPS64 architecture, with working clock rate of 900 MHz;
System buses: VME 2eSST, RapidIO x4, PCI Express x4;
Into composition of the software of units should enter:
The special software for the organisation of performance of functional tasks in real time;
The complex of programs of support of fault tolerance;
The complex of programs of maintenance;
Implementation of a regime of parallel scalings with outcome matching in breakpoints;
Automatic steering of a pattern of units at failure;
Together with functional tasks – conducting in a background of functional check of units.
Possibility of creation and usage of a computer of high efficiency all over the world refers to to factors of strategic potential of defensive and scientific and technical value.
Multiprocessor БЦВМ represent the parallel architecture from several units-servers. In this alternative there is an architecture with "variable access time to memory" (NUMA-architecture) in which memory is physically divided, but is logically readily available. Thus access time to various memories becomes unequal.
In pattern AMP, the operating system is fulfilled on each unit-server, or a multiprocessing system kernel. Each combination модуль/OС is actually the single-processor computer with the rights.
High-speed buses PCI Express x 4 ensure a high-speed interchanging between units-servers (2 GB/C). The analogous pattern of the architecture can use and in БЦВМ on fig. 4.
Thus access time to memories of other units-servers more than in БЦВМ with NUMA architecture. Especially it is appreciable at small sizes of data exchange for one session.
Possibility of creation and usage of a computer of high fault tolerance all over the world too refers to factors of strategic potential of defensive and scientific and technical value.
Apex of a line fault-tolerant БЦВМ are the systems ensuring continuous readiness. The yield with continuous readiness if it works correctly, eliminates any down time both planned, and not planned.
БЦВМ with such performances use for complexes of durable stand - alone maintenance. For example: space objects, type of the coherent companions.
Fault-tolerant БЦВМ are under construction by a principle: троированная the architecture with мажоритированием outcomes of scalings by a rule «2 from 3», the doubled access to input\output devices, power supplies redundant system, operational system with mechanisms реконфигураци and informational recovery.
Reconfiguration fault-tolerant БЦВМ is fulfilled under control of OS according to the pattern table in which the state of each unit-server is mirrored, and the order of exclusion of the defective unit from a pattern is defined. Reconfiguration is fulfilled at the initiative of the principal unit, it delivers the information on a current pattern to all remaining units in БЦВМ. If the principal unit there is a voting fails, and the new principal unit which controls reconfiguration is defined.
Offered architectures of construction of multinuclear units-servers ensure construction as structures high-duty БЦВМ, and construction of structures fault-tolerant БЦВМ constant readiness for solution big a circle of tasks in perspectiv transportable radioelectronic complexes of link and handle, airborne systems of air defence and РЭБ.
- Golovkin B.A. Parallel'ny'e vy'chislitel'ny'e
sistemy'. M.: Nauka. 1980.
- Super E'VM. Apparatnaya i programmnaya
organizacziya / pod red. S. Fernbaxa. M.: Radio i svyaz'. 1991.
- Rasulova S.S. Konspekt lekczij po disczipline
«Nadezhnost' informaczionny'x sistem». 2008.
- Snyder L. A Taxonomy of Synchronous Parallel
Machines. University Park. Penn. 1988.
- Johnson E.E. Completing an MIMD
Multiprocessor Taxonomy // Computer Architecture News. 1988. V. 16. № 2. P. 44–48.
- NetLogic Microsystems,
Multi-Core, Multi-Thread Superscalar Communication Processor XLP832.