JPH07152697A - Roughly coupled computer system - Google Patents

Abstract

PURPOSE:To constitute the roughly coupled computer system which prevents an abnormal packet, etc., from being outputted to a computer coupling network and is improved in trouble resistance not only by starting substitute operation, but also by disconnecting an in-use system which is brought into trouble from the computer coupling network. CONSTITUTION:The computer coupling network 110 of the roughly coupled computer system is equipped with a means 111 which previously stores information showing the correspondence between a logical node number as an identifier specifying a reception-destination computer and the physical node number of a computer determined on the basis of actual connection relation and determines a corresponding physical node number on the basis of the logical node number specified from a transmission-source computer by using the stored information and a means 113 which performs a transfer process to the reception-destination computer having the determined physical node number.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for interconnecting computer systems and a computer system constructed by interconnecting the computer systems, and more particularly to a loosely coupled computer system.

[0002]

2. Description of the Related Art Conventionally, an FTC (Fault) that performs non-stop operation by completely duplicating each resource and automatically switching when a failure is detected for the purpose of surely completing a requested process within a certain period of time. In order to reduce Trelant Computer) and downtime, there is a hot standby configuration that duplicates the entire system and switches to the standby system when a failure occurs. The FTC and hot standby configurations have the advantage that they can maintain the same processing capacity as before the failure even when a failure occurs, but the spare system is a completely redundant part and requires additional resources only to improve reliability. There was a problem of high cost.

Therefore, as a technique for avoiding the entire system down when a failure occurs as compared with that before the failure occurs, a system in which computers are loosely coupled to each other and backups are mutually carried out. Is proposed.

When the backups are mutually exchanged by loose coupling, it is very important to reduce the processing to computers other than the faulty system in order to switch to the standby system at high speed. If the computer that was in communication can continue to request processing for the same node number as before the occurrence of the failure, there is no need to notify and switch to the computer that was in communication with the failed system. It is advantageous.

To achieve this, Japanese Patent Laid-Open No. 1963/1993
As disclosed in Japanese Patent Publication No. 51-51, there is proposed a method of starting an alternative operation including an alternative reception for the same node number used by the failure system before the occurrence of the failure, in accordance with preset information. There is.

Another example of the above method is the document IBM AIXpert.
As shown in 1992 February 39-45, by preparing a plurality of computer-coupled network adapters for alternate reception and enabling alternate reception by the computer node number of the failure side at the time of failure, high-speed operation is possible. The standby system can be switched.

[0007]

In the prior art, the process for the packet for the active system is performed after the failure is detected on the side of the preliminary operation. However, it is possible that the active system continues incomplete operation only by starting the alternative operation, and in that case, an abnormal packet may be output to the computer-coupled network. Increased traffic due to such abnormal packets flowing through the computer-coupled network, increased overhead for the node that received the abnormal packets to perform fault processing, and a network when the protocol for handling abnormal packets is insufficient. There is a problem such as a failure. Therefore, there is a problem that the method according to the conventional technique cannot provide sufficient reliability.

An object of the present invention is to prevent abnormal packets from being output to the computer-coupled network by not only starting the alternative operation but also disconnecting the faulty working system from the computer-coupled network, thereby improving fault tolerance. It is to construct an improved computer-connected network and a loosely-coupled computer system.

[0009]

The configuration of a loosely coupled computer system which is a preferred embodiment of the present invention is shown below. A plurality of computers in the network are connected to each other by a computer connection network. The computer in the network is composed of a processor, a memory, an I / O device, a fault notification line adapter, and a computer connection network adapter connected to the system bus.

The computer connection network is a connection network for connecting a plurality of computers in the network for mutual communication, and is composed of a data transfer control unit and a data transfer network unit. In the data transfer control unit, by holding the logical node number for designating the receiving side computer in the network as the only table that corresponds to the physical node number for each computer in the network for the computer-connected network, the start of alternative operation The failed computer in the network can be separated from the computer connection network.

[0011]

According to the present invention, it is possible to disconnect the working system that has failed at the start of the alternative communication, so that it is possible to prevent abnormal packets from flowing in the computer-coupled network.
A system with high fault tolerance can be realized.

[0012]

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 shows one of the embodiments. The loosely coupled computer system is
The computers in the network (100-1 to 100-N) are connected to each other by the computer connection network 110 and configured. In-network computer (100-1 to 100-N)
Is a processor 101, a memory 102, an I / O device 10
3. The fault notification line adapter 104 and the computer connection network adapter 120 are connected to the system bus 105.
The processor 101 can be configured by a CPU, for example.
The memory 102 can be configured by a RAM, for example. The I / O device 103 is a device for inputting / outputting, and is configured by, for example, a display device, a keyboard, a secondary storage device such as a hard disk. Failure notification line adapter 1
Reference numeral 04 denotes a line for notifying a failure, which can be configured by adding another computer coupling network adapter 120, for example.

The fault notification line adapter 104 is a computer in the network (100-100) depending on the fault detection and notification method.
1 to 100-N) may not be included. I
The / O device 103 may not be included as an element of the computers (100-1 to 100-N) in the network when the I / O device 103 is configured to substitute the input / output function by the communication processing by the computer-coupled network adapter 120.

The computer-coupled network adapter 120 is a part that couples the computers (100-1 to 100-N) in the network and the computer-coupled network 110 to each other, and performs data transmission / reception processing according to an instruction from the processor 101. A control unit 121, a transmission / reception buffer 122, and a computer connection network transmission / reception unit 123 are connected to an internal bus 124.

The computer connection network 110 is a connection network for connecting the computers (100-1 to 100-N) in the network for mutual communication, and comprises a data transfer control unit 111 and a data transfer network unit 113. It

Each in-network computer (100-1 to 100-N) connected to the computer-coupled network 110 designates a connection position, a physical node number statically defined by a switch, etc., and a receiver for communication. It has a logical node number used as an identifier.

Communication performed via the computer coupling network 110 is performed by transmitting and receiving packets. The packet structure is shown in FIG. All packets are the destination node number 2
10, source node number 220, type 230, and data 240. Destination node number 210
Is an identifier that specifies a computer in the network that should receive the packet. The source node number 220 is an identifier indicating the computer in the network that has transmitted the packet. The type 230 is information such as the type and size of the information held by the packet. The data 240 is data to be transferred to the in-network computer designated by the destination node number.

Computers in each network for communication
The destination node number 210 when specifying (100-1 to 100-N) uses the logical node number of each in-network computer (100-1 to 100-N). The computer connection network 110 converts a logical node number into a corresponding physical node number and performs a transfer process.

The data transfer network 113 receives the destination physical node number and the packet from the data transfer control unit 111, and transfers them to the in-network computers (100-1 to 100-N) having the corresponding physical node numbers. This is a part to be performed and can be configured by logic such as TTL or CMOS.

The data transfer control unit 111 is a unit for performing data transfer control by converting a logical node number of transfer data into a physical node number and outputting the physical node number and a packet to the data transfer network 113. The transfer destination determining unit 112, a logical / physical node number conversion information holding unit 114, and a packet buffer 115 are included.

The transfer destination determining unit 112 is a unit that determines the corresponding physical node number from the logical node number in the packet using the contents of the logical / physical node number conversion information holding means 114 and controls the data transfer. , For example, TT
It can be configured by logic such as L or CMOS.

Logical / physical node number conversion information holding means 1
14 is a computer in each network (100-1 to 100
-N) is a part that holds the information on the correspondence between the physical node numbers and the logical node numbers as a single table, and can be configured by logic such as TTL or CMOS or RAM. logic
The contents of the / physical node number conversion information holding means 114 can be rewritten from the computers (100-1 to 100-N) in the network, and are stored as the only table.
By rewriting the contents of the physical node number conversion information holding unit 114, it is possible to change the physical node number corresponding to the logical node number and disconnect the faulty node from the computer-coupled network 110.

The packet buffer 115 is a part for holding packets from the computers (100-1 to 100-N) in each network, for example, logic such as TTL or CMOS or R.
Can be configured by AM.

FIG. 3 shows an embodiment of the logical / physical node number conversion information in the logical / physical node number conversion information holding means 114. As shown in FIG. 3, the physical node number can be searched from the logical node number, and the information can be held as the only table. In this example, there is one correspondence between the physical node number and the logical node number, but there may be multiple correspondences.
If there is more than one, the packet will be transferred to more than one physical node. In this example, N is a positive integer indicating the total number of nodes, and J is a positive integer not less than 1 and not more than N.

The packet transfer control process of the transfer destination determining unit 112 will be described with reference to FIG. First, as shown in step 410, computers in the network (100-1 to 100-
The packet output by N) is taken into the packet buffer 115. Then, as shown in step 420, the logical / physical node number conversion information holding means 11 has the table structure shown in FIG. 3 with the destination node number of the packet fetched in the packet buffer 115 as the logical node number.
4 to determine the corresponding physical node number. Next, as shown in step 430, the physical node number of the search result is notified to the data transfer network unit 113. Finally, as shown in step 440, when the packet fetched in the packet buffer 115 is output to the data transfer network unit 113, the data transfer network unit 113 outputs the packet to the corresponding physical node number.

Next, the processing for performing the alternative operation will be described. First, in this embodiment, it is assumed that the configuration definition between the computers that perform the alternative operation is set in the secondary storage device of the I / O device 103 of the alternative computer using a utility program or the like. That is, when a failure occurs in a certain network computer, the configuration definition between these computers is set in the alternate computer so that another specific computer can perform preliminary processing as the alternate computer.

Hereinafter, a substitution target setting process will be described as an initialization process for the substitution side computer to perform a substitution operation according to FIG. First, as shown in step 510, the I / O device 10
The substitution target information preset in 3 is read out. Next, as shown in step 520, a preliminary program for performing an alternative operation is activated. Here, as shown in step 520, in the alternative computer, the alternative program (spare program) is put to sleep before the initialization of the communication mechanism for the computer-coupled network. Next step 5
As shown in 30, the fault monitoring communication process is activated.
The fault monitoring communication process is, for example, a process in which a non-surveillance computer continues to transmit that there is no fault at regular time intervals, and a substitute computer (monitoring computer) continues to receive the communication to detect a fault.

The fault monitoring communication methods include a notification method using the failure notification line adapter 104 and a notification method using the computer-coupled network without using the failure notification line adapter 104. The former monitoring communication adapter 104
It can be said that the communication method using is more effective than the latter notification method. This is because the former communication method directly notifies the failure between the non-monitoring side computer and the alternative side computer (monitoring side computer), so it is possible to detect the fault more quickly.
This is because the alternative operation can be performed quickly. On the other hand, like the latter communication method, if communication for fault detection is performed in a computer-coupled network, the frequency of fault detection communication (non-monitoring) (The frequency at which the local computer reports that there is no failure at regular intervals)
Therefore, the failure detection time may be delayed.

The alternative operation start processing in the alternative computer after the failure detection will be described below with reference to FIG. First, as shown in step 610, a failure is detected in the failure monitoring communication. Next, according to step 620, the processor 101 reads the number of the alternative node (the node number of the non-monitoring computer (failure computer)) that should perform the alternative operation from the replacement target information read from the I / O device 103. . Next, according to step 630, the processor 101
Make settings to receive the packet with the alternate node number.

At this time, the information held as the only table in the logical / physical node number conversion information holding means 114 of the computer-coupled network 110 is switched to change the node number (non-value) of the computer in the network to be replaced. Setting to set the physical node number of the local node to the physical node number corresponding to the logical node number indicated by the node number of the monitoring-side computer (failure-side computer), and to disconnect the non-monitoring-side computer (failure-side computer) You can After this processing, the packet output to the substitution-side computer that is the substitution target can be substitution-received by the operation described in FIG. Moreover, since the computer on the failure side is disconnected from the computer connection network 110, no abnormal packet is issued.

Next, at step 640, the communication mechanism of the substitute program for performing the substitute operation is activated to start the substitute processing.

In the above embodiment, the configuration definition information for rewriting the table of the logical / physical node number conversion information holding means 114 is used as a utility in the secondary storage device of the I / O device 103 of the alternative computer. It was set using a program. As another example, in the table of the logical / physical node number conversion information holding unit 114, information on a new association between the logical node number and the physical node number of the failed computer and the alternative computer in the event of a failure It is also possible to configure so that the alternative computer can detect the failure and notify the computer connection network 110 only the activation of the table rewriting of the logical / physical node number conversion information holding means 114. is there.

[0033]

As described above, according to the present invention, it is possible to disconnect a working system that has failed at the start of alternative communication. Therefore, by preventing abnormal packets from flowing through the computer-coupled network, fault tolerance is high. The system can be realized.

[Brief description of drawings]

FIG. 1 is an example of a loosely coupled computer system configuration diagram.

FIG. 2 is a diagram showing an example of a computer-connected network packet configuration.

FIG. 3 is an example of logical / physical node number conversion information.

FIG. 4 is a flowchart showing a transfer destination determining unit process.

FIG. 5 is a flowchart showing an alternative target setting process.

FIG. 6 is a flowchart showing an alternative system starting process.

[Explanation of symbols]

100-1 to 100-N: Computer in network, 101:
Processor, 102: Memory, 103: I / O device, 104:
Failure notification line, 105: system bus, 110: computer connection network, 111: data transfer control unit, 112: transfer destination determination unit, 113: data transfer network unit, 114: logical / physical node number conversion information holding means, 115: Packet buffer, 1
20: Computer connection network adapter, 121: Communication control unit, 12
2: Transmission / reception buffer, 123: Computer-coupled network transmission / reception unit,
124: Internal bus.

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Shoichi Murase 1-280, Higashi Koikekubo, Kokubunji, Tokyo Inside Central Research Laboratory, Hitachi, Ltd. (72) Inventor Takeshi Aimoto 1-280, Higashi Koikeku, Kokubunji, Tokyo Hitachi, Ltd. Central Research Laboratory (72) Inventor Hidenori Inouchi 1-280, Higashi Koikekubo, Kokubunji, Tokyo Hitachi Central Research Laboratory (72) Inventor Takehisa Hayashi 1-280, Higashi Koikeku, Tokyo Kokubunji City Central Research Laboratory, Hitachi, Ltd. ( 72) Inventor Hiroshi Iwamoto 810 Shimoimaizumi, Ebina, Kanagawa Prefecture Office Systems Division, Hitachi, Ltd. (72) Toru Horimoto 5030 Totsuka-cho, Totsuka-ku, Yokohama, Kanagawa Prefectural Corporation Hitachi Software Development Division

Claims (6)

[Claims] 1. In a loosely coupled computer system in which a plurality of computers are loosely coupled to each other using a computer coupled network, it is determined from a logical node number which is an identifier for designating a destination computer and an actual connection relationship. Information for indicating correspondence with physical node numbers of the plurality of computers is stored in advance, and means for determining a corresponding physical node number using the information based on the logical node number designated by the transmission source computer. A loosely coupled computer system comprising: a means for performing a transfer process to a destination computer having the determined physical node number, at least in the computer coupled network. 2. In the loosely coupled computer system according to claim 1, when a failure occurs in any of the plurality of computers, the failure is detected by a previously designated alternative computer,
A loosely coupled computer system characterized by rewriting the above information and disconnecting a computer in which a failure has occurred from the computer coupled network. 3. The loosely coupled computer system according to claim 2, wherein among the plurality of computers, a predetermined computer and an alternative computer are connected by a fault notification line different from the computer coupled network, and the alternative computer is A loosely coupled computer system characterized in that a failure occurrence is detected through the failure notification line. 4. A loosely coupled computer system in which a plurality of computers are loosely coupled to each other using a computer coupled network, wherein the plurality of computers include a processor, a memory, and an I / O device including a secondary storage device. , Fault notification line adapter,
The computer-coupled network has a computer-coupled network adapter, and the computer-coupled network has a packet buffer for temporarily storing packets from the source computer, a logical node number that is an identifier for designating the destination computer, and an actual connection relationship. Based on the logical / physical node number conversion table that stores in advance information indicating the correspondence with the determined physical node numbers of the plurality of computers and the logical node number specified by the transmission source computer in the packet, the logical / Transfer destination determining means for determining the corresponding physical node number using the information in the physical node number conversion table, and data transfer for performing the transfer processing of the packet temporarily stored in the destination computer having the determined physical node number Network means, and is connected to the plurality of computers by the computer-coupled network adapter. The computer and the alternative computer are connected to each other via the fault notification line adapter via a fault notification line different from the computer coupling network, and the alternative computer detects a fault occurrence through the fault notification line, A loosely coupled computer system characterized by rewriting the information in the logical / physical node number conversion table of the computer coupled network so as to disconnect the failed computer from the computer coupled network. 5. The loosely coupled computer system according to claim 4, wherein in the packet, the logical node number of the destination computer, the logical node number of the source computer, and the type of information held by the packet or A loosely coupled computer system including type information for determining a size and data to be transferred. 6. The loosely coupled computer system according to claim 4, wherein a substitute program for performing a substitute operation is stored in the secondary storage device of the I / O device of the substitute computer. Combined computer system.