What Is a Multimachine?

A computer system consisting of two or more electronic computers. Connected by two or more computers through shared main memory resources (information) or through data links. According to the different tightness of each computer in the system, it can be divided into two categories: tight multi-machine system and loose disaster multi-machine system, which are generally configured at the same location and do not require communication systems to connect. Failure of any one of the computers does not affect the normal operation of the entire system. The purpose of establishing a multi-machine system is to improve reliability and computing speed. This kind of system has the characteristics of large processing power, high response speed, good reliability, high performance-price ratio, flexible implementation of multiple configurations, etc., and has a wide range of applications.
Multi- computer system

Chinese name: Computer multicomputer system
Foreign name: Multi-computer machine system

Subjects: Electronics, computer, communications
Use: Multi-user operation

Development of computer multi-computer system

A multi-computer system is a computer system composed of more than two electronic computers. It is generally located in the same place and does not require a communication system to connect. Failure of any one of the computers does not affect the normal operation of the entire system. The purpose of establishing a multi-machine system is to improve reliability and computing speed

sample graph Multiprocessors and multiprocessor systems, distributed processing systems and computer networks: Multiprocessors and multiprocessor systems are the only way to further develop parallel technology, and are the main development direction of giant and mainframe. They are multiple instruction stream multiple data stream (MIMD) systems. Each machine processes its own instruction stream (process), communicates with each other, and jointly solves large-scale problems. They have a higher level of parallelism than parallel processors, with great potential and flexibility. Using a large number of inexpensive microcomputers to construct a system through an interconnected network to achieve high performance is one of the directions for studying multiprocessors and multicomputer systems. Multiprocessors and multi-machine systems require parallel algorithms to be studied at higher levels (processes). High-level programming languages provide a means of concurrent and synchronous processes. The operating system is also very complicated, and communication and synchronization between multiple processes must be resolved. , Control and other issues.

A distributed system is a development of a multi-machine system. It is a system that is composed of a plurality of independent but interacting single machines that are physically distributed to coordinately solve user problems. Its system software is more complicated (see distributed computer systems).

Modern mainframes are almost multi-function systems with distributed functions. In addition to high-speed CPUs, there are I / O processors (or front-end user machines) that manage input and output, communication control processors that manage remote terminals and network communications, and Maintenance diagnosis machine for system maintenance diagnosis and database processor for database management. This is a low-level form of a distributed system.

Multiple geographically distributed computer systems are connected to each other through communication lines and network protocols to form a computer network. It is divided into local (local) computer networks and remote computer networks according to the geographical distance. Computers on the network can share information resources and software and hardware resources with each other. Ticket booking systems and information retrieval systems are examples of computer network applications.

Working principle of computer multi-computer system

The general process of a user using a computer system to solve a problem:

Establish an account through the system operator and obtain the right to use it. The account number is used to identify and protect the user's files (programs and data), and the system automatically counts the user's use of resources (accounting, payment).

According to the problem to be solved, study the algorithm, select the appropriate language, write the source program, and provide the data to be processed and related control information.

Put the result of into the floppy disk on the offline dedicated device, and create the user file (also can be done on the online terminal, and the file is directly created in the auxiliary storage, and the fourth step is omitted at this time).

Enter the user files on the floppy disk into the computer with the help of the floppy disk machine, after processing and processing, as a job, register and store in the auxiliary storage.

is required to compile. The operating system calls the job into the main memory, and calls the compiler of the selected language, compiles and links (including the called subroutines), generates a machine-executable target program, and stores it in the auxiliary memory.

requires arithmetic processing. The operating system calls the target program into the main memory, which is processed by the central processing unit, and the result is stored in the auxiliary memory.

The operation result is sent to the external device for output by the operating system in the format required by the user.

The internal work of the computer ( ) is a complicated process under the control of the operating system. Usually, there are multiple user jobs in a computer input at the same time, and they are uniformly scheduled by the operating system and run staggered. But this scheduling is transparent to the user, and the average user does not need to know its internal details. Users can use a terminal to interactively control the operation of (time-sharing mode); they can also entrust the operator to complete , where are automatically performed by the computer (batch mode). The batch processing method is highly automated, but the user is not intuitive and there is no intermediate intervention. The time-sharing method is intuitively controlled by the user and can intervene in error correction at any time, but with a low degree of automation. Most modern computer systems provide two methods, which are selected by the user.

General characteristics of computer multi-computer system

A multicomputer computer has three characteristics:

Computer multi-computer system Modularity

The resource elements of the computer system of the multi-machine system form relatively independent modules, which form a single system through interconnections of the interconnection network. The addition and replacement of modules within a certain range will not affect the integrity of the system.

Computer multi-computer system

Decentralized resource elements can work together to solve a common problem. Under the control of a distributed operating system, different forms of parallelism such as resource duplication (by task) or time overlap (by function) can be achieved.

Computer multi-computer system

The operation of system resources is highly autonomous. There is no master-slave control relationship of the entire system, and the principle of processing localization can be used to reduce the amount of data communication between nodes.

The generation and development of computer systems for multi-machine systems is driven by a number of factors. The first is technical factors: large-scale integrated circuits and microprocessors provide inexpensive hardware for multi-machine computer systems; the development of digital communication technologies and computer network technologies has enabled a large number of computer nodes to be interconnected and communicate at high speeds. become possible. More important is the user factor. Although the centralized time-sharing computer system can connect a large number of remote and short-range terminals to meet the needs of geographically dispersed multiple users, the excessive concentration of processing functions will cause problems such as large communication overhead, long response time, and complex and expensive systems. Users are gradually turning to distributed computer systems in order to seek higher technical and economic benefits.

Multi-computer system Computer system has become a new direction of rapid development in the field of computer research. It also has a series of features due to its comprehensive distribution characteristics.

Potential advantages of computer multi-computer systems

Computer multi-computer system Reliability and ruggedness

Resource redundancy and autonomous control enable the system to be dynamically reconfigured and continue to work even after partial damage. The modularity of the system facilitates maintenance and use.

Computer multi-computer system Incremental scalability

Inexpensive modules are used as system expansion or resource update increments, and the entire system does not have to be replaced like a centralized system.

Computer multi-computer system Flexibility

The configuration of the system is easy to change to meet the various needs of different application objects.

Computer multi-computer system

Computer resources are closer to the user, especially to enable scattered small users to get fast response and direct services from the computer, thus combining the computing power of the mainframe with the convenience of a microcomputer.

Computer multi-computer system Resource sharing

Sharing software and hardware resources on the basis of being transparent to users will make the available resources of a single user multiply.

Computer multi-computer system

Parallel processing capabilities distributed by task are less limited by the size of the system; dedicated processing components distributed by function can also enhance the effective processing capabilities of the system.

Computer Multi-Computer SystemEconomy

Conducive to the performance of price advantage of microcomputers.

Computer multi-computer system: adapt to various application environments

The resource allocation at each node of the distributed computer system can be well matched with the needs of local users, so it is especially suitable for economic management, transaction management, process control and other applications that have scattered users and require mutual coordination.

Computer multi-computer system master-slave

A master-slave operating system records, controls the status of other slave processors, and assigns tasks to the slave processors by a master processor. For example, Cyber-170 is a master-slave multiprocessor operating system. It resides on a peripheral processor Po, and all other processors, including the central processor, are subordinate to Po. Another example is DECSystem10, which has two processors, one master and one slave. The operating system runs on the main processor, and the request from the slave processor is transmitted to the main processor through trapping, and then the main processor answers and performs the corresponding service operation. The monitoring program of the master-slave operating system and the process of providing services do not have to be migrated because only the master processor utilizes them. When an unrecoverable error occurs, the system can easily cause a crash, and the main processor must be restarted. Due to the heavy responsibility of the master processor, when it is too late to process the process request, the utilization of other slave processors will decrease accordingly.

The master-slave operating system has the following characteristics:

A. The operating system program runs on a processor. If the slave processor needs the master processor to provide the service, it sends a request to the master processor, and the master processor accepts the request and provides the service. It is not necessary to write the entire hypervisor as reentrant program code, because only one processor is using it, but some common routines must be reentrant.

B. Since only one processor accesses the execution table, there are no management table access conflicts and access blocking issues.

C. When the main processor fails, it is easy to cause the entire system to crash. If the main processor is not a fixed design, the administrator can choose one of the other processors as the new main processor and restart the system.

D. Task allocation is not only easy to make part of the slave processor idle and cause system efficiency to decline.

E. For asymmetrical systems where the workload is not too heavy or consists of processors with very different functions.

F. The system consists of a master processor plus several slave processors. The hardware and software structure is relatively simple, but flexible.

Independent supervision of computer multi-computer system

Separate supervisor is different from master-slave. In this type, each processor has its own management program (core). Multiprocessor systems using independent supervised operating systems include IBM370 / 158.

Independent supervision features:

A. Each processor will perform various management functions according to its own needs and the tasks assigned to it. This is called independence.

B. Since there are several processors executing the hypervisor, the code of the hypervisor must be reentrant, or a dedicated copy of the hypervisor must be loaded for each processor.

C. Because each processor has its own dedicated management program, there are fewer conflicts in accessing public forms, naturally there are fewer blocking situations, and the efficiency of the system is high. But conflict arbitration agencies are still needed.

D. Each process is relatively independent, so a failure of one processor will not cause the entire system to crash. However, it is very difficult to remedy the damage caused by the failure or re-execute the unfinished work of the failed machine.

E. Each processor has dedicated I / O devices and files.

F. This type of operating system is suitable for a loosely coupled multiprocessor system, because each processor has a local memory to store a copy of the management program, the storage is too redundant, and the utilization rate is not high.

G. Independently supervised operating systems are more difficult to achieve processor load balancing.

Computer multi-computer system floating supervision

Floating supervisor (floating supervisor) has only one processor at a time as the "main processor" that performs comprehensive management functions, but according to the needs, the "main processor" is floating, that is, switching from one processor to another . This is the most complex, efficient, and flexible multiprocessor operating system, and is often used in symmetric multiprocessor systems (that is, the permissions of all processors in the system are the same, with common main memory and I / O subsystems) . The floating supervisory operating system is suitable for tightly coupled multiprocessor systems. Multiprocessor systems using this operating system include MVS running on IBM3081, VM, Hydra running on C · mmp, and so on.

Features of floating supervision:

A. There is only one processor at a time as the "main processor" that performs comprehensive management functions, but allows multiple processors to execute the same management service subroutine at the same time. Therefore, most hypervisor code must be reentrant.

B. According to need, the "main processor" is floating, that is, switching from one processor to another. In this way, even if the main processor executing the management function fails, the system can still operate as usual.

C. Some non-specialized operations (such as I / O interrupts) can be sent to those processors that are least busy during a certain period of time to make the system's load reach a better balance.

D. Service request conflicts can be resolved by priority methods, and access conflicts on shared resources are resolved by mutually exclusive methods.

E. The processors in the system are managed by the concept of processor sets, where each processor can be used to control any I / O device and access any memory block. This management method is transparent to the processor, and has high reliability and considerable flexibility.

Application of computer multi-machine system computer multi-machine communication in substation

In the substation, the real-time detection of its various parameters is very important, and the workload is also very large. In the past, manual recording methods were used, and problems such as misrememberment and omission were often encountered. For substations in remote areas, it was impossible to achieve various Real-time detection of parameters. With the continuous development of computer communication technology, it is possible to solve the above problems. The substation detection system has been put into practical use. In this system, there is an unattended operation between the single-chip microcomputer, the single-chip microcomputer and the computer, and the substation. Automatic remote communication ^[1] .

System Components

The system consists of three levels: the front-end machine, which is composed of AT89c2051 single-chip microcomputer, which mainly completes the real-time collection of watt-hour meter values, voltage, and current. Each energy-metering unit requires a front-end machine; the management machine, which consists of 8031 single-chip microcomputers, mainly It manages up to 255 front-end machines, including collecting various data collected by front-end machines, and can place the front-end machine into the frequency division coefficient of the pulse electric meter through the keyboard on the management machine, observe the collected data, Modify the date, time, etc., and send the collected substation data to the upper computer and receive the time correction data sent by the upper computer; the upper computer, which mainly completes the report summary, printing, recording, querying, and remote management of the computer End control, sending data in the local area network ^[1] .

working principle

The cutting-edge machine and the management machine are placed in the substation. They are installed in the control cabinet and fixed by the instrument panel. Each group of control screens needs a management machine. Each management machine can detect 255 front-end machines. The management machine is at the level. In a multi-machine system, to ensure the reliability of the communication between the master and the slave, the communication interface must be capable of identifying, and the control bit in the serial control register SCON in 8031 SM2 is set up to meet this requirement. When the serial port works in mode 2 (mode 3), each frame of information transmitted and received is 11 bits, of which the 9th data bit is a programmable bit. The TBS of SCON is assigned 1 or 0 to distinguish whether an address frame or a data frame is sent. If the slave control bit SM2 = 1, then when an address frame is received, the data is loaded into SBUF and R1 = 1 is set to the CPU. Send an interrupt request; if a data frame is received, no interrupt flag will be generated and the information will be discarded. If SMZ = 0, both the address frame and the data frame will generate an R1 = 1 interrupt flag, and the data will be loaded into SBUF. According to these Requirements, the communication process between the master and the slave is: The slave SM2 bit is set to 1 and it is only receiving address frames. The master sends a frame of address information, which contains an 8-bit address, and the 9th bit is l to indicate that the address is sent. After the slave receives the address frame, Compare the received address with its own address. The addressed slaves clear their SM2. The other slaves that are not addressed still maintain SM2 = 1, and the master sends data or control information (the 9th bit is 0). For the slaves that have been addressed, because SMZ = 0, they can receive the information sent by the master. For other slaves, because the SMZ remains at 1, the data frames sent by the master will be ignored. Until a new address frame is sent. When the master contacts another slave, it can send another address frame to address its slave, and the previously addressed slave analyzes that the master is addressing other slaves. At this time, it recovers its SM2 = 1, and ignores the subsequent data frames sent by the host ^[1] .