CN106302700B - Distributed system consistent update method and system based on paxos algorithm - Google Patents

Abstract

The invention discloses a consistent update method and system of a distributed server system based on paxos algorithm, comprising: a master node sends master node state information to a slave node; and the slave node judges whether the state of the master node is the latest according to the state information of the master node state; if not, then send the slave node’s own slave node status information to the master node; if so, set the master node status information as the status information of the slave node’s own node; wherein, the master node status information and the slave node status information are both It is the state information carrying the time label; the master node determines the latest state information from all slave node state information, and sets the latest state information as the current state information of the master node, and sends the current state information to the slave node; the slave node according to the current state information The status information updates the status information of its own node. It can be seen that, through the consistency update solution provided by this embodiment, the states of different nodes in the system can be updated at any time to ensure the consistency of the system.

Description

Distributed system consistency update method and system based on paxos algorithm

technical field

The present invention relates to the field of server software, more specifically, to a method and system for consistent update of a distributed server system based on paxos algorithm.

Background technique

Distributed server system is a configuration method of computer hardware and corresponding function configuration method. It is a multi-controller computer system, and each controller forms a unified system through an interconnected network. The system adopts a distributed computing structure, that is, the tasks processed by the central processor in the original system are distributed to the corresponding controllers, so that the controllers with different functions can coordinate with each other and share the peripherals and software of the system. From the outside, the entire server system is unified and transparent. For each node controller, it is necessary to ensure their consistent state. The initial state of each node controller is consistent, and multiple controllers need to communicate during the working process to ensure consistency. Distributed node communication models mainly include shared memory communication model and message passing communication model. Among them, in the shared memory model, multiple processes share a certain section of physical memory, and the consistency is maintained by the minority obeying the majority. Fault tolerance is not strong enough. The message passing model has strong fault tolerance and can be applied to abnormal situations such as message delay, loss, and even network hangup. The distributed server system speeds up the processing speed of the system and simplifies the logical structure of the host computer. It is especially suitable for the automatic control of industrial production lines and the management of enterprises and institutions. It has low cost and is easy to maintain. It has become an important direction for the development of computers in the application field.

In a single-controller database, users modify data in the database without worrying about other users modifying the same data at the same time. However, in a multi-controller database, statements in multiple transactions executing concurrently can modify the same data. Concurrently executing transactions need to produce meaningful and consistent results. Therefore, in a multi-controller database, how to maintain data consistency is a problem to be solved by those skilled in the art.

Contents of the invention

The object of the present invention is to provide a method and system for updating consistency of a distributed server system based on paxos algorithm, so as to maintain data consistency.

In order to achieve the above object, the embodiment of the present invention provides the following technical solutions:

A consistent update method for a distributed server system based on the Paxos algorithm, comprising:

The master node sends master node status information to the slave node;

The slave node judges whether the state of the master node is the latest state according to the master node status information; if not, then sends the slave node status information of the slave node itself to the master node; It is set as the state information of the slave node itself; wherein, the master node state information and the slave node state information are state information carrying a time stamp;

The master node determines the latest state information from all slave node state information, and sets the latest state information as the current state information of the master node, and sends the current state information to the slave nodes;

The slave node updates the state information of its own node according to the current state information.

Wherein, the slave node judges whether the state of the master node is the latest state according to the state information of the master node, including:

The slave node compares the time stamp in the state information of the master node with the time stamp of the state information stored by itself;

If the timestamp in the status information of the master node is the latest, it is determined that the status of the master node is the latest state; if the timestamp in the status information of the master node is not the latest, it is determined that the status of the master node is not the latest state.

Wherein, the master node determines the latest status information from all slave node status information, including:

The master node obtains corresponding time labels from all slave node state information;

The slave node state information corresponding to the latest time label is taken as the latest state information.

Among them, after the master node sends the master node status information to the slave node, it also includes:

Start the timer to start timing;

When the timing time of the timer reaches the predetermined duration, detect whether the slave node status information sent by the slave node is received;

If received, execute the step of the master node determining the latest status information from all slave node status information; if not received, then determine that the master node status information is the latest status information.

Wherein, the sending the current state information to the slave node includes:

After the master node sets the latest state information as the current state information of the master node, it uses the current state information of the master node as the master node state information, and sends it to the slave node in the next period.

A consistent update system of a distributed server system based on the paxos algorithm, including a master node and a slave node;

The master nodes include:

The first sending module is used to send the master node status information to the slave node;

The current state information determination module is used to determine the latest state information from all slave node state information, and set the latest state information as the current state information of the master node, and trigger the sending module to send the slave node the Describe the current status information;

The slave nodes include:

A judging module, configured to judge whether the state of the master node is the latest state according to the state information of the master node; if not, trigger the second sending module; if so, trigger the setting module;

The second sending module is configured to send slave node status information of the slave node itself to the master node;

The setting module is used to set the state information of the master node as the state information of its own node of the slave node, and update the state information of its own node according to the current state information; wherein, the state information of the master node and the state information of the slave node The state information is state information carrying a time stamp.

Wherein, the judging module compares the time stamp in the state information of the master node with the time stamp of the state information stored in the slave node itself; if the time stamp in the state information of the master node is the latest, then judge the state of the master node is the latest state; if the time stamp in the state information of the master node is not the latest, then it is determined that the state of the master node is not the latest state.

Wherein, the current state information determination module obtains corresponding time labels from all slave node state information, and uses the slave node state information corresponding to the latest time label as the latest state information.

Wherein, the master node also includes:

Timing module, used to start the timer to start counting;

The detection module is used to detect whether the slave node status information sent by the slave node is received when the timing time of the timer reaches a predetermined duration; if received, the master node determines the latest status from all slave node status information information step; if not received, then determine that the master node state information is the latest state information.

Wherein, after the current state information determination module sets the latest state information as the current state information of the master node, it uses the current state information of the master node as the master node state information, and transmits One cycle is sent to the slave node.

It can be seen from the above scheme that a consistent update method and system of a distributed server system based on paxos algorithm provided by the embodiment of the present invention includes: the master node sends the master node status information to the slave node; information to determine whether the state of the master node is the latest state; if not, then send the slave node status information of the slave node itself to the master node; if so, set the master node status information as the slave node’s own node State information; wherein, the state information of the master node and the state information of the slave node are state information carrying time labels; the master node determines the latest state information from all the state information of the slave nodes, and sets the latest state information is the current state information of the master node, and sends the current state information to the slave node; the slave node updates the state information of its own node according to the current state information. It can be seen that, through the consistency update solution provided by this embodiment, the states of different nodes in the system can be updated at any time to ensure the consistency of the system.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a schematic flow chart of a consistent update method for a distributed server system based on the paxos algorithm disclosed in an embodiment of the present invention;

Fig. 2 is a schematic diagram of the boss broadcasting its own state information disclosed in the embodiment of the present invention;

Fig. 3 is a schematic diagram of the common reply information disclosed by the embodiment of the present invention;

Fig. 4 is a schematic diagram of the status of the boss update system disclosed in the embodiment of the present invention;

FIG. 5 is a structural block diagram of a consistency update system of a distributed server system based on the paxos algorithm disclosed in an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

The embodiment of the invention discloses a consistent update method and system of a distributed server system based on a paxos algorithm, so as to maintain data consistency.

Referring to Fig. 1, a kind of consistency update method of the distributed server system based on paxos algorithm provided by the embodiment of the present invention, comprises:

S101. The master node sends master node status information to the slave node;

Among them, after the master node sends the master node status information to the slave node, it also includes:

Start the timer to start timing;

When the timing time of the timer reaches the predetermined duration, detect whether the slave node status information sent by the slave node is received;

If received, execute the step of the master node determining the latest status information from all slave node status information; if not received, then determine that the master node status information is the latest status information.

Specifically, this embodiment is a distributed system consistency design of an improved paxos algorithm. In a distributed server system, multiple controller nodes are divided into two roles: master node boss and slave node common. At any time, there is only one master node boss, and the remaining controller is the slave node common. In the initial state, each controller node in the distributed system is in the initial state. Each controller node in the system can receive external messages and interact with the external world. Each controller node reacts to change the state according to the received message, and updates its own state information meg at any time. The state information meg contains a time stamp marked with the time of receiving the message. In this way, after a period of time, the state of the controller node in the system will be differentiated. Therefore, in this solution, in order to ensure that the state of the entire distributed system is always consistent and updated at any time, the boss will update the system state uniformly.

Specifically, in this embodiment, the only boss in the distributed system regularly broadcasts its state information meg_boss to multiple commons, and the information includes a time stamp. Referring to Fig. 2, in a lease, the boss controller broadcasts its own state information meg_boss to the common controller in the system, and the state information meg_boss has a time stamp.

S102. The slave node judges whether the status of the master node is the latest state according to the status information of the master node; if not, send the status information of the slave node itself to the master node; The state information is set as the state information of the slave node itself; wherein, the master node state information and the slave node state information are state information carrying time labels;

Wherein, the slave node judges whether the state of the master node is the latest state according to the state information of the master node, including:

The slave node compares the time stamp in the state information of the master node with the time stamp of the state information stored by itself;

If the timestamp in the status information of the master node is the latest, it is determined that the status of the master node is the latest state; if the timestamp in the status information of the master node is not the latest, it is determined that the status of the master node is not the latest state.

Specifically, for a single common, it will compare the time stamp of the state information meg_boss received with the state information meg time stamp stored by itself. If the status message meg_boss received by common is not up-to-date compared to the status message meg stored by itself, then the meg stored by itself will be sent to boss; if meg_boss is up-to-date relative to common’s meg, common will use meg_boss instead to update the meg stored by itself , do not return information to the boss.

Specifically as shown in Figure 3, the common controller responds to the received meg_boss accordingly. Among them, the state message meg_boss received by commonnode1 is the latest compared to the state message meg stored by itself, so meg_boss is used to replace and update the meg_1 stored by itself, and no information is returned to the boss. The meg_boss received by common node2 is not up-to-date relative to its own meg_2, so it sends meg_2 to the boss. The meg_boss received by common node3 is not up-to-date compared to its own meg_3, and it also sends meg_3 to the boss.

S103. The master node determines the latest state information from the state information of all slave nodes, sets the latest state information as the current state information of the master node, and sends the current state information to the slave nodes;

Wherein, the master node determines the latest status information from all slave node status information, including:

The master node obtains corresponding time labels from all slave node state information;

The slave node state information corresponding to the latest time label is taken as the latest state information.

Specifically, referring to Fig. 4, the boss controller will compare the received meg_2 with meg_3, and since meg_3 is relatively new, the boss controller selects meg_3 to update its own state information. As shown in Figure 4, in the next lease, the boss controller broadcasts the latest status information meg_3 to the common controllers in the system; each common controller updates the status again according to meg_3.

S104. The slave node updates the state information of its own node according to the current state information.

Wherein, the sending the current state information to the slave node includes:

After the master node sets the latest state information as the current state information of the master node, it uses the current state information of the master node as the master node state information, and sends it to the slave node in the next period.

Specifically, if the boss receives the reply message from the common within the specified time lease, then among the meg replies from multiple commons, the latest content is determined as the state information meg_boss of the boss. If the lease expires and the boss does not receive a reply message from any common, it means that all the controllers in the system are updated to the latest status meg_boss. In the next lease period, the boss broadcasts its state information meg_boss to the common in the system again. In this way, the entire system is guaranteed to be updated and consistent at any time.

Specifically, in this solution, after the master node sends the master node status information to the slave nodes, a timer is started to start timing, where the timing time of the timer may be the above-mentioned specified time lease. Then in this scheme, the time to send the current status information of the master node to the slave node again is the time when the master node sends the status information to the slave node again in the next cycle.

The consistency update system provided by the embodiment of the present invention is introduced below, and the consistency update system described below and the consistency update method described above may refer to each other.

Referring to FIG. 5 , a consistent update system of a distributed server system based on the paxos algorithm provided by an embodiment of the present invention includes a master node 100 and a slave node 200;

The master nodes consist of 100:

The first sending module 101 is configured to send the master node status information to the slave node;

The current state information determination module 102 is used to determine the latest state information from all slave node state information, and set the latest state information as the current state information of the master node, and trigger the sending module to send the slave node said current state information;

The slave node 200 includes:

Judging module 201, configured to judge whether the state of the master node is the latest state according to the state information of the master node; if not, trigger the second sending module; if so, trigger the setting module;

The second sending module 202 is configured to send the slave node status information of the slave node itself to the master node;

The setting module 203 is used to set the state information of the master node as the state information of its own node of the slave node, and update the state information of its own node according to the current state information; wherein, the state information of the master node and the state information of the slave node The node state information is state information carrying a time stamp.

Based on the above technical solution, the judging module compares the time stamp in the state information of the master node with the time stamp of the state information stored in the slave node itself; if the time stamp in the state information of the master node is the latest, then it is determined that the master The state of the node is the latest state; if the time stamp in the state information of the master node is not the latest, it is determined that the state of the master node is not the latest state.

Based on the above technical solution, the current state information determination module obtains corresponding time labels from all slave node state information, and uses the slave node state information corresponding to the latest time label as the latest state information.

Based on the above technical solution, the master node also includes:

Timing module, used to start the timer to start counting;

The detection module is used to detect whether the slave node status information sent by the slave node is received when the timing time of the timer reaches a predetermined duration; if received, the master node determines the latest status from all slave node status information information step; if not received, then determine that the master node state information is the latest state information.

Based on the above technical solution, after the current state information determination module sets the latest state information as the current state information of the master node, it uses the current state information of the master node as the master node state information, and through the The sending module sends to the slave node in the next cycle.

A consistent update method and system for a distributed server system based on the paxos algorithm provided by an embodiment of the present invention includes: the master node sends master node status information to the slave node; the slave node judges the master node status information according to the master node status information Whether the status of the node is the latest status; if not, then send the slave node status information of the slave node itself to the master node; if so, then set the master node status information as the status information of the slave node itself; wherein, Both the master node status information and the slave node status information are status information carrying time labels; the master node determines the latest status information from all slave node status information, and sets the latest status information as the master node and send the current state information to the slave node; the slave node updates the state information of its own node according to the current state information. It can be seen that, through the consistency update solution provided by this embodiment, the states of different nodes in the system can be updated at any time to ensure the consistency of the system.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. a kind of consistent update method of the distributed server system based on paxos algorithm characterized by comprising

Host node sends primary node status information to from node；

Judge whether the state of the host node is last state from node according to the primary node status information；If it is not, then to The host node sends the slave node status information from node itself；If so, by the primary node status information be set as from The status information of node own node；Wherein, the primary node status information and from node status information be carry the time mark Number status information；

The host node determines newest status information from all from node status information, and the newest status information is set For the current state information of the host node, and the current state information is sent to from node；

The status information of own node is updated according to the current state information from node.

2. consistent update method according to claim 1, which is characterized in that it is described from node according to the host node shape State information judges whether the state of the host node is last state, comprising:

The Time labeling of the status information stored from node according to the Time labeling in primary node status information with itself is compared Compared with；

If Time labeling in primary node status information be it is newest, determine the state of the host node for last state；If Time labeling in primary node status information be not it is newest, then determine that the state of the host node is not last state.

3. consistent update method according to claim 1, which is characterized in that the host node is from all from node state Newest status information is determined in information, comprising:

The host node obtains corresponding Time labeling from all from node status information；

Using slave node status information corresponding to newest Time labeling as newest status information.

4. consistent update method described in any one of -3 according to claim 1, which is characterized in that host node is to from node After transmission primary node status information, further includes:

Starting timer starts timing；

After the timing time of timer reaches scheduled duration, detect whether to receive the slave node state letter sent from node Breath；

If receiving, the host node is executed from all the step of determining newest status information from node status information； If being not received by, determine that the primary node status information is newest status information.

5. consistent update method according to claim 4, which is characterized in that described to send the current shape to from node State information, comprising:

After the newest status information is set as the current state information of the host node by the host node, by the host node Current state information sent as primary node status information, and in next cycle to from node.

6. a kind of consistent update system of the distributed server system based on paxos algorithm, which is characterized in that including main section Point and from node；

The host node includes:

First sending module, for sending primary node status information to from node；

It is described to include: from node

Judgment module, for judging whether the state of the host node is last state according to the primary node status information；If It is no, then trigger the second sending module；If so, trigger setup module；

Second sending module, for sending the slave node status information from node itself to the host node；

The host node includes:

Current state information determining module, for determining newest status information from node status information from all, and will be described Newest status information is set as the current state information of the host node, and triggers first sending module and send to from node The current state information；

It is described to include: from node

The setup module, for setting the status information from node own node, and root for the primary node status information The status information of own node is updated according to the current state information；Wherein, the primary node status information and from node state Information is to carry the status information of Time labeling.

7. consistent update system according to claim 6, which is characterized in that

The judgment module according in primary node status information Time labeling with from node itself store status information when Between label be compared；If Time labeling in primary node status information be it is newest, determine that the state of the host node is Last state；If the Time labeling in primary node status information be not it is newest, determine that the state of the host node is not most New state.

8. consistent update system according to claim 6, which is characterized in that

The current state information determining module obtains corresponding Time labeling from all from node status information, and will be newest Time labeling corresponding to slave node status information as newest status information.

9. the consistent update system according to any one of claim 6-8, which is characterized in that the host node also wraps It includes:

Timing module starts timing for starting timer；

Detection module is received for detecting whether after the timing time of timer reaches scheduled duration from node transmission From node status information；If receiving, the host node is executed from all and determines last state from node status information The step of information；If being not received by, determine that the primary node status information is newest status information.

10. consistent update system according to claim 9, which is characterized in that the current state information determining module The newest status information is set as after the current state information of the host node, the current state of the host node is believed Breath is used as primary node status information, and is sent in next cycle to from node by first sending module.