TWI629882B - A method for achieving a stable selection strategy by a coordinator of distributed communication based on JGroups - Google Patents

Abstract

The invention discloses a coordinator's stable selection strategy, which is suitable for a distributed communication system based on JGroups, and can provide multiple selection strategies according to the applicable environment, and dynamically selects the best members according to the strategy in the process of group member change. Serve as a coordinator. At the same time, we also provide a mechanism to enable the coordinator to change quickly without affecting the original communication work.

Description

A method for achieving a stable selection strategy for coordinators of decentralized communication based on JGroups

The present invention pertains to a coordinator by having a member in a distributed communication group that is responsible for maintaining and publishing the latest status of members to all other members of the group.

Taking a US patent Combined join and state transfer, patent number: US8312084 B2 as an example, the patent focuses on the state synchronization or transfer maintenance mechanism when the node joins, but does not target whether the coordinator role needs to be transferred after the node is joined. How to transfer the method.

If the coordinator unfortunately suffers from inefficiency due to itself or certain environmental factors, but because the role cannot be transferred, this phenomenon will continue to improve and cannot be improved.

In view of the shortcomings derived from the above-mentioned conventional methods, the inventors of the present invention have improved and innovated, and after years of painstaking research, they finally succeeded in researching and developing the method for the stable selection strategy of the coordinators based on JGroups to achieve decentralized communication.

In order to achieve the above object, the present invention proposes to provide a program based language JGroups: A method for a coordinator of decentralized communications to securely select a strategy, including: Step 1. A new member asks to join a group or the original coordinator leaves the group; Step 2. Trigger a selection procedure; Step 3. If in In a high-load environment, all the members of the new member are reported to the coordinator host for their performance data. Step 4. If it is in a high-traffic environment, all members including the new member are calculated to the RTT of each member, according to The average value is sorted; Step 5. If the preference is set by itself, the host is prioritized according to the configuration file; Step 6. After the selection process is completed, the new coordinator is selected according to the sorting result; Step 7. The new coordinator issues the final status result Give all other members.

The high-load environment is to comprehensively evaluate the CPU usage, memory usage, and disk usage to effectively stabilize the group communication mechanism; and to select the lowest-loaded member of the group member as the new coordinator.

The high-traffic environment refers to the average of a member's packet to the back-to-back time of all other members in a period of time to effectively stabilize the group communication mechanism, and to select the group operator with the lowest average RTT as the new coordinator.

When the selection process is triggered, the four selection strategies are simultaneously proposed, and the user can select an appropriate strategy according to different environmental characteristics. The default is the original mechanism of JGroups. First, the longest survival time, that is, the first member to join the group, the coordinator will not change after that; Second, according to the host performance, when the setting is at high load (High Loading) In a decentralized environment, the present invention evaluates the computational efficiency of all members according to the settings, and then selects one with the lowest load as the new coordinator; third, according to the average RTT response time, when the setting is in the high flow (High Flow) In a decentralized environment, the present invention observes the average of each member to the Round Trip Time of all other members over a period of time. Select one of the lowest average RTT as the new coordinator; Fourth, according to the system configuration settings, the user can write the host priority in the mind to the configuration profile in advance, the system will automatically select the best according to the configuration content order. A role as a coordinator.

In the design mechanism of JGroups, the coordinator has more responsibility for maintaining the latest status of the group than other members, and also shoulders the responsibility of synchronizing group information to other members. The terminology of group information in JGroups is called View. What are the current members of the group, who is the group coordinator, and the latest status of each member. The coordinator will periodically send the latest View to each member to synchronize the latest group information. The coordinator of the original JGroups mechanism is fixed. The first member who joins the group is the coordinator. Then the B members, C members, and D members join, but the coordinator will not change regardless of the group members. However, because member A is the first to join the group, it means that the group time is the longest. In general, it can be assumed that it is relatively stable. In addition, in the heartbeat program that detects the immediate state of all members, the coordinator must also perform the final confirmation action. Once a member fails to detect the neighboring member, the coordinator will immediately report the abnormality, and the coordinator will confirm the details. To determine if an abnormal member is really offline.

The invention provides a method for a stable selection strategy of a coordinator based on JGroups to achieve decentralized communication, and has the following advantages when compared with other conventional technologies:

1. In a highly loaded, highly loaded environment, the present invention automatically selects a member with the lowest load as the coordinator.

2. In a high-flow distributed environment, the present invention automatically selects a member with the smallest Round Trip Time as the coordinator.

3. Coordinators can be selected in order of user-selected order to prioritize stable members.

4. The coordinator will not affect the communication work of the original group during the entire re-selection process.

S110~S170‧‧‧ Process

210‧‧‧A member

Member of 211‧‧‧B

212‧‧‧C members

Member of 213‧‧‧D

220‧‧‧Group Information

The detailed description of the present invention and the accompanying drawings will be further understood, and the technical contents of the present invention and the functions thereof can be further understood. The related drawings are: FIG. 1 is a stable selection strategy of a coordinator who achieves distributed communication based on JGroups. Flowchart of the method; FIG. 2A, FIG. 2B, FIG. 2C and FIG. 2D are schematic diagrams of the method for achieving a stable selection strategy of the coordinator of the distributed communication based on the JGroups.

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Hereinafter, the present invention will be further described with reference to the accompanying drawings:

Please refer to FIG. 1 , which is a flowchart of a method for implementing a stable selection strategy for a coordinator of distributed communication based on JGroups, including: Step 1. S110 new member requests to join a group or the original coordinator leaves the group; Step 2 S120 triggers the selection process; step 3. If it is in a high load environment, S130 includes all the members of the new member to report to the coordinator host's respective performance data and then sorts; Step 4. If it is in a high traffic environment, S140 includes new All members of the member are calculated to the RTT of each member and sorted according to the average value; Step 5. If the preference is set by itself, the S150 performs host prioritization according to the configuration file; Step 6. After the selection process is completed, the S160 selects a new coordinator according to the sorting result; Step 7. The S170 new coordinator issues the final status result to other All members.

The high-load environment is to comprehensively evaluate the CPU usage, memory usage, and disk usage to effectively stabilize the group communication mechanism; and to select the lowest-loaded member of the group member as the new coordinator.

The high-traffic environment refers to the average of a member's packet to the back-to-back time of all other members in a period of time to effectively stabilize the group communication mechanism, and to select the group operator with the lowest average RTT as the new coordinator.

In summary, there are two trigger timings for the coordinator selection process, when new members are asked to join the group, because the new members may be more suitable as coordinators, or the original coordinator leaves the group. If the coordinator is gone, it is natural to choose a new coordinator. If the former new member requests to join the group, no matter which member receives the request from the new member to join the group, it will be forwarded to the coordinator, and the coordinator will trigger the selection process. If the original coordinator leaves the group, the member responsible for detecting the coordinator finds that the coordinator is no longer in the group, and naturally no coordinator can convey it. This member will trigger the selection process. Regardless of the trigger method, we call the member of the trigger selection program a trigger member, and the trigger member may be any member of the group.

The triggering member will first review the environment before deciding on the selection strategy. When setting up a highly loaded, decentralized environment, the triggering member will ask the group members to perform a self-efficacy assessment process that includes CPU Utilization, Memory Utilization, and disk usage. (Disk Utilization) comprehensive assessment, and get a set of evaluation scores and return Trigger member.

When setting up a high-traffic, decentralized environment, the triggering member will ask the group members to perform a self-RTT averaging time calculation program. The program will observe the average time after all other members' packet round-trip time and convert it into a set of evaluation scores. After the return to the trigger member.

When the setting is in the system priority setting according to the system configuration profile, the triggering member does not need the information of other group members, and can be converted into a group evaluation score according to the order position of each online group member on the setting file.

If none is set, the default is the original mechanism of JGroups, that is, the first member to join the group is the new coordinator, but we still generate a set of evaluation scores, but obviously the highest score for the first group. The second high score, even if it is the first member offline, will not affect the subsequent selection judgment.

Some options may be time consuming, such as in the least average RTT time, but no matter which one, before the selection is not over, the trigger member will temporarily work for the coordinator and post this message to all members of the group. In order to prevent the original group communication mechanism from shutting down. And in the collection of the assessment information of the group members, the principle of Best Efforts is adopted, and if the return is not returned within the unit time, the candidate qualification is excluded.

The triggering member will sort the evaluation scores based on the collected evaluation scores, select the new coordinator, and trigger the member to inform the new coordinator of the message. Note that the new coordinator may also trigger the member himself, and the new coordinator will be the first. After the release of the latest group information, other members were also aware of the new coordinator.

When the selection process is triggered, the four selection strategies are simultaneously proposed, and the user can select an appropriate strategy according to different environmental characteristics. The default is the original mechanism of JGroups. First, the longest survival time, that is, the first member to join the group. Since then, the coordinator has not changed; Second, according to the host performance, when setting the high load (High Loading) decentralized environment, the present invention will evaluate the computing performance of all members according to the settings, and then select a load from it. The lowest as the new coordinator; Third, according to the average RTT response time, when setting the high-flow (High Flow) decentralized environment, the present invention will observe the packet round-trip time of each member to all other members in a period of time (Round Trip Time), after averaging, select one of the lowest average RTT as the new coordinator; Fourth, according to the system configuration settings, the user can write the host priority in the mind to the configuration profile in advance, the system will follow the group The order of the content content automatically selects the best one to act as a coordinator. However, depending on the host performance and the average RTT response time, it may take more time. Therefore, the selection process proposed by the present invention automatically picks up the role of a member temporary coordinator, so that during the coordinator change process, Affect the progress of the original communication work.

Please refer to 2A, 2B, 2C, 2D, which is a schematic diagram of the method for achieving a stable selection strategy of the coordinator of distributed communication based on JGroups. In the design mechanism of JGroups, the coordinator is responsible for maintenance compared with other members. The latest state of the group work, but also shoulders the responsibility of synchronizing group information to other members. The terminology of group information 220 in JGroups is called View (Group Information 220), which records which members and groups of the current group. Who is the coordinator and the latest status of each member. The coordinator will periodically send the latest View 220 to each member to synchronize the latest group information. The coordinator of the original JGroups mechanism is fixed. The first member A who joins the group is selected as the coordinator, and then the B member 211, the C member 212, and the D member 213 are successively joined, but regardless of how the group members change, the coordinator Neither will change, but because A member 210 is the first to join the group, meaning that the group time is the longest, in general, it can be assumed to be relatively stable. In addition, in detecting the real-time status of all members In the Heartbeat program, the coordinator must also perform the final confirmation action. Once a member does not detect the neighboring member, the coordinator will immediately report the abnormality. The coordinator will confirm the confirmation to determine whether the abnormal member is offline. However, if Heartbeat returns an abnormality due to environmental conditions, a network partition will occur.

If Heartbeat returns an abnormality due to environmental conditions, Network Partition will occur. Please refer to 2C and 2D to illustrate the original mechanism in the face of this problem, group members and coordinators. The process of change. As shown in FIG. 2C, it is assumed that the C member 212 loses contact with the original group due to a network abnormality, and the B member 211 will report to the coordinator A member 210 because the C member 212 is not detected, and the coordinator confirms and confirms again. The C member 212 is really offline, so the C member 212 is deleted from the group information, and the D member 213 is designated as the new neighbor member of the B member, and the B member 211 is changed to the state of the probe D member 213, and then the coordinator will be the latest group. After the information is released to the B and D members, it indicates that the C member 212 has completed the separation process. C member 212 has no connection with the original group members, but has to stand on his own as a new group, and since it is the only member of the new group, it is undoubtedly the natural coordinator of the new group. Figure 2D follows the scenario of Figure 2C, assuming that the D member 213 does not contact the original group, but is in contact with the C member 212. Soon, the B member 211 finds that the D member 213 is not detected, and the D member 213 is also kicked out of the communication group along the same previous steps, so that the original communication group has only the A member 210 and the B member 211, and each other. Do Heartbeat. The split D member 213 is able to communicate with the C member 212, so joins the new communication group formed by the C member 212, and the coordinator still selects the first split C member 212. Therefore, looking at the above, it can be observed that the coordinator has greater demands on computing power and network traffic than other members. But in the process of changing members of the group, they are unable to adapt to the situation. Make better choices.

The detailed description of the present invention is intended to be illustrative of a preferred embodiment of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

To sum up, this case is not only innovative in terms of technical thinking, but also has many of the above-mentioned functions that are not in the traditional methods of the past. It has fully complied with the statutory invention patent requirements of novelty and progressiveness, and applied for it according to law. Approved this invention patent application, in order to invent invention, to the sense of virtue.

Claims (5)

A method for a stable selection strategy based on JGroups to achieve decentralized communication, including: step 1. new members request to join the group or the original coordinator to leave the group; step 2. trigger the selection process; step 2-1. In a high-load environment, all members of the new member are reported to the coordinator host's respective performance data and then sorted; Step 2-2. If it is in a high-traffic environment, all members including the new member are calculated to each member's RTT. Sorting according to the average value; Step 2-3. If the preference is set by itself, the host is prioritized according to the configuration file; Step 3. After the selection process is completed, the new coordinator is selected according to the sorting result, wherein the selection program, Including four selection strategies, it is: the longest survival time, is the first member to join the group, since then the coordinator no longer changes; according to the host performance, when setting a high-load decentralized environment After evaluating the performance of all members according to the settings, select one with the lowest load as the new coordinator; according to the average RTT response time, when the setting is at high flow In the high-flow decentralized environment, it will observe the average time between each member and all other members after the packet is round-trip time, and select one with the smallest average RTT as the new coordinator; according to the system configuration settings, The user can write the priority of the host in the configuration configuration file in advance, and the system automatically selects the best role as the coordinator according to the configuration content order; Step 4. The new coordinator issues the final status result to All other members. The method for stably selecting a strategy for a coordinator based on JGroups to achieve decentralized communication, as described in claim 1, wherein the high load environment is a comprehensive evaluation including CPU usage, memory usage, and disk usage. To effectively stabilize the group communication mechanism. The method for solidifying the selection strategy by the coordinator of the distributed communication based on JGroups, as described in claim 1, wherein the high-load environment is to select the member with the lowest load among the group members as the new coordinator. The method for stably selecting a strategy for a coordinator who has reached a decentralized communication based on JGroups as described in claim 1, wherein the high-traffic environment refers to an average of a member's packet to and from all other members in a period of time. To effectively stabilize the group communication mechanism. The method for solidifying the selection strategy by the coordinator of the distributed communication based on JGroups as described in claim 1 of the patent scope, wherein the high-traffic environment is to select the new average coordinator among the group members as the new RTT.