JPH06348660A - Distributed processing network - Google Patents

Abstract

PURPOSE:To detect an error of the reference time of a time server and to adopt the correct time in a decentralized processing network. CONSTITUTION:A client node gives a request to a time server to transmit the reference time in a power supply application state and then stores the received time in a network time register 9. Besides, the time of a clock circuit 15 at this time is stored in a time register 10. Then the absolute value of the difference of time between both registers 9 and 10 is stored in a difference detecting register 12 and then compared with the value of an allowable value setting part 13 by a difference comparator 14. If the difference absolute value is smaller than the allowable value, the time of the register 9 is loaded in the circuit 15. Meanwhile the present time of the circuit 15 and the time of the register 9 are shown on a display 7 when the difference absolute value is larger than the allowable value. So that an operator chooses that he adapts whether either one of the time of the register 9 or 10, or another corrected time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed processing network in which time is synchronized between nodes (computers).

[0002]

2. Description of the Related Art In processing such as file management in a distributed processing network, even if a plurality of transactions (processing) are simultaneously executed in parallel, the order of reading and writing to a file is as if they were processed in order. Need to manage. This is called transaction serialization.

As a method for realizing the serialization of transactions, a method based on a time stamp is generally used. The method gives a time stamp to each transaction at the start of the transaction, and when multiple transactions compete for a certain file and request access, the transactions are serialized by allowing access only in the order of the time stamp. It is about to be transformed. In order to serialize transactions by this time stamp-based method, it is essential to synchronize the time between nodes.

As a method for synchronizing the time between nodes, conventionally, there is a method in which a time supply transmission line is provided between the time supply device and each node and the time is supplied from the time supply device to each node. (Eg, Japanese Patent Laid-Open No. 4-77841)
However, a special transmission line is required, which is disadvantageous in terms of cost. On the other hand, as a time synchronization method that does not require a special transmission line such as a time supply transmission line, at least one node on the distributed processing network is used as a time server, and the time of each node is based on the time of the time server. A method of calibrating is known.

Next, the method will be described. Figure 4
FIG. 1 is a block diagram showing an outline of a distributed processing network. 4, 20 and 30 are LANs (local area networks), 21 is a print server, 21a is a printer control unit, 22 is a file server, 22a is a file management unit, 22b and 31b are resource management units, and 23 and 32 are communications. Servers 23a and 32a are communication control units 24 and 2.
5, 33 and 34 are workstations, 31 is a mail server, and 31a is a mail management unit.

Each server and workstation (hereinafter referred to as a node) in such a distributed processing network usually has its own internal clock circuit, but these clock circuits always provide accurate time. It does not always output. Therefore, at least one of those nodes has a time server function, and the other nodes (hereinafter, referred to as client nodes) set the time of their own clock circuit based on the time output from the node. Calibrate. For example, the client node inquires the time server about the current time when the power is turned on, and adjusts the time of its own clock circuit to the time returned.
When a plurality of time servers are provided, one main time server is determined and the other time servers match their respective times with the main time server.

A conventional document relating to such a distributed processing network is, for example, Japanese Patent Laid-Open No. 2-15.
6758, JP-A-2-186415 and JP-A-4-97409.

[0008]

However, in the above-mentioned conventional technique in which the time of each node is calibrated based on the time of the time server, when the time server causes a time shift, the time There is a problem in that the node that receives the signal cannot detect it and calibrates the time of its own clock circuit based on the shifted time. An object of the present invention is to solve such a problem by making it possible to detect the time shift of the time server.

[0009]

In order to solve the above problems, according to the present invention, at least one time server for transmitting a reference time and a client for receiving the reference time and calibrating the time of its own clock circuit. In a distributed processing network having a node,
The received reference time is used only when the absolute value of the difference between the received reference time and the time of its own clock circuit is within the allowable value.

[0010]

[Operation] The time server sends the reference time in response to the request from the client node, and when the client server receives the reference time from the time server, it compares the received reference time with the time of its own clock circuit. Then, the reference time received from the time server is used only when the absolute value of the difference is within the allowable value. Therefore, even if the time server causes a time shift, the client node that receives the time detects that there is a time shift and does not use it, so the time of its own clock circuit is calibrated based on the time shift. It will not happen.

[0011]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an outline of a client node in a distributed processing network. Figure 1
, 1 is a CPU, 2 is a memory, 3 is a LAN controller, 4 is a disk controller, 5 is a fixed disk, 6
Is a display controller, 7 is a display, 8 is a network address register, 9 is a network time register, 10 is a time register, 11 is a comparison circuit, 1
2 is a difference detection register, 13 is an allowable value setting unit, 14 is a difference comparison circuit, 15 is a clock circuit, 16 is a crystal oscillation circuit, 1
Reference numeral 7 is an error register.

The client node requests the time server to transmit the reference time and stores the received time in the network time register 9. On the other hand, the time of the clock circuit 15 at that time is stored in the time register 10. Then, the absolute value of the difference between the time of the network time register 9 and the time of the time register 10 is stored in the difference detection register 12, and that value and the value of the allowable value setting unit 13 are compared by the difference comparison circuit 14, When the absolute value of the difference is less than or equal to the allowable value, the clock circuit 15 is loaded with the time of the network time register 9. However, when the absolute value of the difference exceeds the allowable value, the current time of the clock circuit 15 and the current time of the network time register 9 are displayed on the display 7, and which time should be adopted for the operator. , Or use other correction time. In addition,
In the network address register 8, when a plurality of time servers are installed on the network, the time server of the reference time transmission source is configured to return error information and the like to the time server of the reference time transmission source. Holds the network address.

Next, the operation when the client node requests the time server to transmit the time and receives the time response packet will be described using a flowchart. FIG. 2 is a flowchart showing a processing procedure when the client node requests the time server to transmit the time and receives the time response packet. Step 1 ... When the power is turned on, the client node requests the time server to transmit the reference time by broadcasting a time request packet as shown in FIG. The time server is
When this time request packet is received, the reference time is read from its own clock circuit, a time response packet as shown in FIG. 3B is constructed, and it is transmitted to the client node which issued the time request packet.

Step 2 ... It is determined whether or not a time response packet has been received. Step 3 ... When received, time information is extracted from the received time response packet and the time is stored in the network time register 9. If the network is an international network, the time information in the time response packet contains the offset azimuth in addition to the current time, the longitude of the place where the time server is installed, and the offset time in the world. Includes the difference from standard time, information about daylight saving time, the time required to transmit a time response packet, etc. Step 4 ... The time of the clock circuit 15 is stored in the time register 10. Step 5: The comparison circuit 11 compares the time of the network time register 9 with the time of the time register 10. Step 6 ... Store the absolute value D of the difference between the two times in the difference detection register 12. Step 7 ... It is determined whether or not the absolute value D of the difference is less than or equal to the allowable value P. Step 8 ... In the following steps, the clock circuit 15 is loaded with the time in the network time register 9.

Step 9 ... If not received in step 2, it is judged whether or not a predetermined time has elapsed. When step 10 ... Has passed, the count value C is incremented. Step 11 ... It is determined whether or not the count value C exceeds a predetermined value K. Step 12 ... When it exceeds, or when the absolute value D of the difference is not less than or equal to the allowable value P in step 7, the error register 1
Set that there was an error in 7. Step 13 ... The current time of the clock circuit 15 and the current time of the network time register 9 are displayed. Step 14 ... Confirms with the operator whether or not the time of the clock circuit 15 should be corrected, and determines which one the operator has selected. Step 15: When the operator instructs not to correct, the time of the clock circuit 15 is used as it is to start the node. Step 16 ... When the operator gives an instruction for correction in step 14, the time of the clock circuit 15 is corrected to the time input by the operator, and then the node is started.

When the error is set in step 12, it is possible to notify the time server of the error by e-mail and correct the time of the time server.

[0017]

As described above, according to the distributed processing network of the present invention, even if the time server causes a time shift, the client node side that receives the time detects that there is a time shift, and detects it. Since it is invalid, the time of its own clock circuit will not be calibrated based on the shifted time.

[Brief description of drawings]

FIG. 1 is a block diagram showing an outline of a client node in a distributed processing network.

FIG. 2 is a flowchart showing a processing procedure when a client node requests a time server to transmit time and receives a time response packet.

FIG. 3 is a diagram showing an outline of a time request packet and a time response packet.

FIG. 4 is a block diagram showing an outline of a distributed processing network.

[Explanation of symbols]

1 ... CPU, 2 ... Memory, 3 ... LAN control unit, 4 ... Disk controller, 5 ... Fixed disk, 6 ... Display controller, 7 ... Display, 8 ... Network address register, 9 ... Network time register,
10 ... Time register, 11 ... Comparison circuit, 12 ... Difference detection register, 13 ... Allowable value setting value, 14 ... Difference comparison circuit,
15 ... Clock circuit, 16 ... Crystal oscillator circuit, 17 ... Error register, 20, 30 ... LAN, 21 ... Print server,
22 ... File server, 23, 32 ... Communication server, 2
4, 25, 33, 34 ... Workstation, 31 ... Mail server

Claims (1)

[Claims] 1. A distributed processing network having at least one time server for transmitting a reference time and a client node for receiving the reference time and calibrating the time of its own clock circuit, wherein the client node receives A distributed processing network, which uses the received reference time only when the absolute value of the difference between the reference time and the time of its own clock circuit is within an allowable value.