WO2011000199A1

WO2011000199A1 - Method and system for intelligently scheduling cluster servers

Info

Publication number: WO2011000199A1
Application number: PCT/CN2009/076279
Authority: WO
Inventors: 汤敏
Original assignee: 深圳市融创天下科技发展有限公司
Priority date: 2009-06-30
Filing date: 2009-12-30
Publication date: 2011-01-06
Also published as: CN101938504A

Abstract

A method and system for intelligently scheduling cluster servers are provided, and the system includes: a group of cluster servers, on which a feedback module is set respectively, for collecting realtime information about operation states of each server; a state database connected with the cluster servers, for recording realtime information about operation states of servers according to the realtime information about operation states of each server collected by the feedback module; an IP address database connected with the state database, in which the IP address of each server and the geographic location and network operator information corresponding to the IP address of the server are recorded; a DNS server connected with the IP address database, for allocating the same domain name to the cluster servers which provides the same service, scoring each server in real time according to the operation states in the state database and IP information of the visitor, obtaining the score level of the servers, and choosing the most suitable server IP to provide network service for users.

Description

Instruction manual

Title of Invention: Method and system for cluster server intelligent scheduling

#细 or

[1] The present invention relates to the field of the Internet, and in particular, to a method and system for intelligent scheduling of a cluster server.

[2] In the case of a bottleneck in the performance of a single server, the cluster system can distribute all requests to multiple node servers through a load balancer, which greatly improves the system processing capability. In addition, the load balancer can monitor the availability of server nodes. Currently, the load balancing policies used by the load balancing scheduler are as follows: 1. Round-robin equalization: Each time a request for processing network data traffic is allocated to the internal server in turn. , from 1 to N and then restart, this equalization algorithm is suitable for all servers in the server group have the same hardware and software configuration and the average service request is relatively balanced; 2. Random equalization: 选择 select a server with a random algorithm 3. Minimum connection number balance: Select the server with the smallest load selected from the currently working server; 4. Processing capacity balance: In the processing capacity balance, each server is assigned a weighting coefficient according to the processing capacity of each server, and the weighting coefficient Multiply the absolute load parameter on the server to obtain the relative load parameter, and balance the processing capability to select a server with the smallest relative load parameter. 5. Dynamic Balance: The so-called dynamic load balancing is based on the cluster server status (main processing part such as CPU and memory). To assign tasks. In the first and second balancing strategies, because the feedback information of the server is not introduced, the balancing strategy cannot be adjusted according to the actual situation. The load of each server will be uneven. If a server suddenly fails, the system cannot know. If the fault occurs, some users will not be able to obtain the service. In the third equalization strategy, although the feedback information of the server is introduced, the processing power of different servers is not considered, and the absolute load is used as the measurement standard, and the processing capability of the server is different. A large service cluster is not effective. The fourth balancing strategy takes into account the different processing capabilities of different servers. Each server is assigned a static weighting factor, and the relative load parameter is introduced as a measure. The effect of load balancing, which is a balancing strategy that we usually use; the fifth balancing strategy feeds back the dynamic load information of the server back to the balancing strategy. The effect of balancing is the best, but it is difficult to implement.

[3] In addition, users of the Internet are distributed throughout the country. According to the current status of the network, there is a distance, north and south. Problems such as the network, how to automatically access the user to a server with the closest geographical location and belonging to the telecom or China Netcom has become a problem we need to consider; the capabilities of each server are uneven, and the same service for multiple users The CPU, memory, and number of remaining service users are different. We need to find the server with the lightest load for the user.

[4] In view of this, it is necessary to propose an improved method to overcome the deficiencies of the prior art.

[5] In view of this, the present invention aims to provide a method and system for intelligent scheduling of cluster servers, so as to ensure that network users are allocated to servers that can be accessed, so that network users can efficiently obtain network services.

[6] In order to solve the above problems, the present invention provides a method for intelligent scheduling of a cluster server, and the specific steps of the method are as follows:

[7] A: Establish a state database, which is used to record the actual information of the server running status.

[8] B: Establish an IP address database, which records the IP address of each server and the corresponding information corresponding to the IP address of the server;

[9] C: Set a feedback module on each server to collect the actual information of the running status of the server, and feedback and update to the status database;

[10] D: Establish a DNS server, including a cluster server IP list, assign the same domain name to the cluster server providing the same service, and perform each server according to the running status in the state database and the IP information of the visitor. Rating, resulting in a server rating rating;

[11] E: The DNS server selects the most appropriate server IP from the cluster server IP list according to the rating level to provide network services to the user.

[12] In addition, the present invention further provides a cluster server intelligent scheduling system, wherein the system comprises:

[13] A group of cluster servers, each of which has a feedback module for collecting actual information about the running status of each server. The actual information includes the CPU usage of the server, the memory usage, the server setting load, and the remaining Number of service users;

[14] A state database connected to the cluster server, based on the various servers collected by the feedback module The actual state of the row state, which is used to record the actual information of the running state of the server;

[15] An IP address database connected to the state database, the IP address database records the IP address of each server and the geographical location of the corresponding server IP address and network operator information, and the network operator information includes telecommunications and Netcom;

[16] A DNS server connected to the IP address database, including a cluster server IP list, assigning the same domain name to the server cluster providing the same service, according to the running status in the state database and the IP information of the visitor. The server scores to obtain the rating level of the server, and selects the most suitable server IP according to the rating level to provide network services for the user.

[17] The present invention has the beneficial effects of collecting the running state set of the server according to the feedback module installed on the server. In addition, the present invention provides the server closest to the geographical location of the visitor to serve the DNS according to the geographical location of the visitor's IP. According to the running status, the server can quickly and effectively select the server to provide services for the visitors, so that the visitor's access can be more securely responded.

Country deletion

[18] The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

[1] FIG. 1 is a flow chart of a cluster server intelligent scheduling method according to the present invention;

[20] FIG. 2 is a schematic structural diagram of a cluster server intelligent scheduling system according to the present invention;

[21] FIG. 3 is a schematic structural diagram of a cluster server intelligent scheduling system according to the present invention.

difficult

[22] The specific implementation of the present invention will be described below with reference to the accompanying drawings.

[23] FIG. 1 is a schematic diagram of a cluster server intelligent scheduling method according to the present invention. As shown in the figure, the method includes

[24] S10: Establish a state database, which is used to record the actual information of the running state of the server;

[25] S20: Establish an IP address database, where the IP address of each server records the corresponding information of the IP address of each server and the IP address of the corresponding server; the corresponding information of the IP address of the corresponding server includes the geographical location of the server and network operation. Business information, network operator information includes telecommunications and Netcom.

[26] S30: Set a feedback module on each server to collect the actual information of the running status of the server, and feedback and update it to the status database. The actual information collected by the feedback module includes the service. The CPU usage of the server, the memory usage, the server setting load, and the number of remaining service users, the feedback module is about 5 seconds to feed back the information and update it to the state database. When one of the servers fails, the feedback module will feed back the fault ID 1 to the status database, and the failed server will be marked as invalid in the cluster server IP list. The user does not participate in the comparison score with other servers; after the troubleshooting, the feedback module will feed back the fault ID 0 to the status database, and the failed server will be re-marked as valid in the cluster server IP list.

[27] S40: Establish a DNS server, including a cluster server IP list, assign the same domain name to the cluster server providing the same service, and perform each server according to the running status in the state database and the IP information of the visitor. Rating, resulting in a server rating rating;

[28] In this step, the DNS server includes an IP automatic query module. After the DNS server receives the visitor's access request, the IP automatic query module will automatically query the geographical location of the visitor's IP. Each network server is scored by the geographic location of the IP, the specific geographic location of each network server, and the network operator information, as well as the actual information of each network server's operational status.

[29] In this step, the DNS server scoring principle sequence is as follows: First, according to the CPU usage, memory usage, server setting load, and remaining service users of the server provided by the feedback module

According to the server setting load amount and the number of remaining service users, the server with the largest number of remaining service users has the highest rating of the first level; the number of remaining service users of the server is lower than the set range, where the remaining number of users is set to 100 units, continue to score according to the CPU usage, and the score of the CPU usage is lower than the second level; the number of remaining service users in the server is lower than the set range, and the CPU usage is higher than the set range, where the CPU is occupied. The rate setting range is 70%, and the score is continued according to the memory usage. The score with less memory usage is the third level; the number of remaining service users in the server reaches a certain range, and the CPU usage is higher than the set range and memory. Above the setting range, where the memory usage setting range is 70%, and then based on the visitor's IP geographic location score, the distance between the visitor's IP geographic location and the server's IP geographic location is recently rated as the fourth level.

[30] S50: The DNS server selects the most suitable server IP from the cluster server IP list according to the rating level to provide network services for the user, and the hierarchical order of the DNS server selection server is: first level, second level, third Level, fourth level.

[31] In order to more specifically explain the method and system of the present invention, please refer to FIG. 2, which shows the cluster service of the present invention. Schematic diagram of the intelligent scheduling system of the server; the system includes:

[32] A group of cluster servers, including server 1, server 2, server _3, and server 4, each of which is provided with a feedback module 1, 2, 3, 4 for collecting each server 1, 2, 3, 4 The actual information of the running status, the actual information includes the CPU usage of the server, the memory usage, the server setting load, and the number of remaining service users;

[33] A state database connected to the group cluster server, based on the actual information of the running status of each server collected by the feedback module, the state database is used to record the actual information of the running state of the server;

[34] An IP address database connected to the state database, the IP address database records the IP address of each server and the geographical location of the corresponding server and network operator information, and the network operator information includes telecommunications and Netcom;

[35] A DNS server connected to the IP address database, including a cluster server IP list, assigning the same domain name to the server cluster providing the same service, according to the running status in the state database and the IP information of the visitor. The server performs scoring. The DNS server further includes an IP automatic query module. After the DNS server receives the visitor's access request, the IP automatic query module will automatically query the geographical location of the visitor's IP. Each web server is scored by the geographical location of the visitor's IP, the specific geographic location of each web server and network operator information, and the actual information of each network server's operational status.

[36] The DNS server scoring principle sequence is as follows: Firstly, according to the CP U occupancy rate, memory occupancy rate, server setting load amount and remaining service users of the server provided by the feedback module, the server sets the load amount and the number of remaining service users. The server score is up to the first level; after the number of remaining service users of the server is lower than the set range, the score is continued according to the CPU usage, and the score with less CPU usage is the second level; the number of remaining service users at the server If the CPU usage is lower than the set range, and the CPU usage is higher than the set range, the score will continue to be scored according to the memory usage. The score with less memory usage is the third level; the remaining service users of the server are lower than the set range. If the CPU usage of the peer is higher than the set range and the memory is higher than the set range, and then according to the IP geographic location of the visitor, the distance between the IP address of the visitor and the IP geographic location of the server is recently scored as the fourth level; The user remaining number is set to 100 units, the CPU usage setting range is 70%, and the memory usage setting range is With a range of 70%, the DNS server selects the server in the order of rank: first level, second level, third level, fourth level. Get service The rating level of the device is based on the rating level to select the most appropriate server IP to provide users with network services.

[37] Embodiment 1

[38] The load of server 1 is 2,000, and the remaining load is 200. The CPU usage is 60% and the memory usage is 71%. The load set by server 2 is 2,200. There are 1 800 users accessing, the remaining load is 400, the CPU usage is 55%, the memory usage is 69%, the load set by server 3 is 1,900, and there are already 1,650 users, and the remaining load is 250. The CPU usage rate is 73% and the memory usage is 71%. The server 4 load is 2,100 units and has 1,800 users. The remaining load is 300, the CPU usage is 64%, and the memory usage is 70%. %.

[39] The above data is fed back to the status database by the feedback modules 1, 2, 3, 4, because the remaining load of the four servers is 200, 400, 250, 300, respectively, which are greater than the set remaining load 100, DNS The CPU usage and memory occupancy data of the servers 1, 2, 3, and 4 will not be considered, and the score level of the server 2 is the first level, and the server 2 is directly selected to provide services for the user.

[40] Example 2

[41] The load of server 1 is 2,000, and there are already 1950 users. The remaining load is 50, the CPU usage is 60%, and the memory usage is 71%. The load set by server 2 is 2,200. There are already 214 0 users accessing, the remaining load is 60, the CPU usage is 55%, the memory usage is 69%, and the load set by the server 3 is 1,900 users have access to 1830 users. The number is 70, the CPU usage is 73%, the memory usage is 71%; and the server 4 load is 2100, which has 2020 users access, the remaining load is 300, the CPU usage is 64%, the memory is occupied. The rate is 70%.

[42] The above data is fed back to the status database by the feedback modules 1, 2, 3, 4, because the remaining load of the four servers is 50, 60, 70, 80, respectively, which are greater than the set remaining load 100, DNS The remaining load of the servers 1, 2, 3, and 4 will not be considered. According to the CPU usage of 60%, 55%, 73%, and 64%, the CPU usage of the server 2 is less than the set value of 70% and the minimum, so The score is the highest, and the score of the server 2 is the second level, and the server 2 is directly selected to provide services for the user.

[43] Example 3

[44] The load of server 1 is 2,000, and there are already 1950 users. The remaining load is 50, the CPU usage is 71%, the memory usage is 67%, and the load set by server 2 is 2,200. There are already 214 0 users accessing, the remaining load is 60, the CPU usage is 72%, the memory usage is 66%, the monthly 艮 The load capacity set by server 3 is 1,900 users have access, the remaining load is 70, the CPU usage is 73%, the memory occupancy is 65%; and the server 4 load is 2100, there are already 2020 User access, the remaining load is 300, the CPU usage is 74%, and the memory usage is 64%.

[45] The above data is fed back to the status database by the feedback modules 1, 2, 3, 4, since the remaining loads of the four servers are 50, 60, 70, 80, respectively, which are greater than the set remaining load 100, and The CPU usage is 71%, 72%, 73%, 74%, respectively, which are higher than the set value of 70%. The DNS will not consider the remaining load and CPU usage of the servers 1, 2, 3, and 4, respectively, according to the memory usage. If the memory usage of the server 4 is less than the set value of 70%, the score is the highest, and the score level of the server 4 is the third level, and the server 4 is directly selected to provide services for the user.

[46] Embodiment 4

[47] The load of server 1 is 2,000, and there are already 1950 users. The remaining load is 50, the CPU usage is 71%, the memory usage is 71%, and the load set by server 2 is 2,200. There are already 214 0 users accessing, the remaining load is 60, the CPU usage is 72%, the memory occupancy is 72%, and the load set by the server 3 is 1,900 users have access to 1830 users, the remaining load The number is 70, the CPU usage is 73%, the memory usage is 73%; and the server 4 load is 2100, which has 2020 users accessing, the remaining load is 300, the CPU usage is 74%, the memory is occupied. The rate is 74%.

[48] The above data is fed back to the state database by the feedback modules 1, 2, 3, 4, since the remaining loads of the four servers are 50, 60, 70, 80, respectively, which are greater than the set remaining load 100, and The CPU and memory usage are 71%, 72%, 73%, and 74%, respectively, which are higher than the set value of 70%. The DNS will not consider the remaining load of the servers 1, 2, 3, and 4 and the CPU and memory usage. According to the geographic location of the visitor IP and the geographical location between the servers, the server closest to the geographical location of the visitor is selected. For example, if the server 2 is closest to the visitor, the server 2 has the highest score, and the score of the server 2 is the fourth rank. Directly select server 2 to provide services to users.

[49] FIG. 3 is a modification of the intelligent scheduling system shown in FIG. 2, wherein the state database and the IP address database are both set in the DNS server.

In addition, an improvement (not shown) of the intelligent scheduling system shown in FIG. 3 further includes one or more standby DNS servers connected to the DN S server, and the data of the DNS server is updated synchronously with the data of the standby DNS server. When the DNS server fails, it continues to work by the alternate DNS server. The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalents, and improvements made within the spirit and scope of the present invention should be included in the present invention. Within the scope of protection of the invention.

Claims

Claim

[Claim 1] 1. A cluster server intelligent scheduling method, comprising the following steps: Step A: establishing a state database, the state database is used to record actual information of a server running state;

Step B: Establish an IP address database, where the IP address of each server records corresponding information corresponding to the IP address of the server;

C: Set a feedback module on each server to collect the actual information of the running status of the server, and feedback and update to the status database;

D: Establish a DNS server, which includes a cluster server IP list, assign the same domain name to the cluster server providing the same service, and perform each server according to the running status in the state database and the IP information of the visitor. Rating, resulting in a server rating rating;

E: The DNS server selects the server IP from the cluster server IP list according to the rating level to provide network services for the user.

[Claim 2] 2. The cluster server intelligent scheduling method according to claim 1, characterized in that

In the step B, the corresponding information corresponding to the IP address of the server includes the geographic location of the server and the network operator information, and the network operator information includes the telecommunications and the network communication.

[Claim 3] 3. The cluster server intelligent scheduling method according to claim 2, characterized in that

The actual information collected by the feedback module in the step C includes the CPU usage of the server, the memory usage, the server setting load, and the number of remaining service users. The feedback module is about 5 seconds to feedback and update the status to the status. In the database.

[Claim 4] 4. The cluster server intelligent scheduling method according to claim 3, characterized in that

In the step D, the DNS server includes an IP automatic query module. After the DNS server receives the visitor's access request, the IP automatic query module will automatically query the geographical location of the visitor's IP, and the DNS access is based on the access. Geographic location of the IP, specific geographic location of each network server and network operator information, and each network server The actual status of the running status scores each web server.

[Claim 5] 5. The cluster server intelligent scheduling method according to claim 4, characterized in that

The priority level of the DNS server is: first level, second level, third level, and fourth level; wherein, according to the CPU usage, memory usage, and server settings of the server provided by the feedback module The load amount and the number of remaining service users are compared. According to the server setting load amount and the number of remaining service users, the server with the largest remaining load scores the first level; after the number of remaining service users of the server is lower than the set range, according to the CPU usage Rate continues to score, the score of less CPU usage is the second level; the number of remaining service users in the server is lower than the set range, and the CPU usage is higher than the set range, and then continues to score according to the memory usage, memory usage The score with low rate is the third level; the number of remaining service users on the server is lower than the set range, the CPU usage is higher than the set range and the memory is higher than the set range, and then the IP rating of the visitor is used. , the distance between the IP location of the visitor and the IP location of the server is recently rated as The fourth level.

[Claim 6] 6. The cluster server intelligent scheduling method according to claim 5, characterized in that

The setting of the number of remaining users is 100, the CPU usage setting range is 70%, the memory usage setting range is 70%, and the ranking order of the DNS server selection server is: Level, second level, third level, fourth level.

[Claim 7] 7. A multi-server intelligent scheduling system, characterized in that: the system comprises: a group of cluster servers, each server is provided with a feedback module for collecting actual information of the running status of each server, The actual information includes the CPU usage of the server, the memory usage, the server setting load, and the number of remaining service users; a state database connected to the group cluster server, according to the actual information of the running status of each server collected by the feedback module, The state database is used to record the actual information of the running state of the server;

An IP address database connected to the state database, the IP address database records the IP address of each server and the geographical location of the corresponding server and network operator information, and the network operator information includes telecommunications and Netcom; A DNS server connected to the IP address database, including a cluster server IP list, assigning the same domain name to the server clusters providing the same service, and performing actual operations for each server according to the running status in the state database and the IP information of the visitor. The score is obtained by the server's rating level, and the server IP is selected according to the rating level to provide the user with network services.

[Claim 8] 8. The multi-server intelligent scheduling system according to claim 7, wherein: the state database and the IP address database are both disposed in a DNS server, and the DN S server further includes an IP automatic Query module, after the DNS server receives the visitor's access request, the IP automatic query module will automatically query the geographical location of the visitor's IP, and the DNS will be based on the geographic location of the visitor's IP and the specific geographic location of each web server. Each web server is scored by location and network operator information and actual information about the operational status of each web server.

[Claim 9] 9. The multi-server intelligent scheduling system according to claim 7, wherein: the rating level of the DNS server is from high to low: first level, second level, third level, and Four levels; among them, according to the CPU usage, memory usage, server setting load, and remaining service users provided by the feedback module, according to the server setting load and the number of remaining service users, the server with the largest number of remaining users The score is the first level; after the number of remaining service users of the server is lower than the set range, the score is continued according to the CPU usage, and the score of the CPU usage is less than the second level; the number of remaining service users on the server is lower than the setting. The range, the same CPU usage is higher than the set range, and then continue to score according to the memory occupancy rate, the memory occupancy rate is less than the third level; the number of remaining service users in the server is lower than the set range, the same CPU When the occupancy rate is higher than the set range and the memory is higher than the set range, the score is based on the visitor's IP location. The distance between the IP address of the IP address of the visitor and the IP address of the server is the fourth highest level; the set number of the remaining user is 100, and the CPU usage setting range is 70%, and the memory usage rate is The setting range is 70%, and the ranking order of the DNS server selection server is: first level, second level, third level, fourth level. [Claim 10] 10. The multi-server intelligent scheduling system according to claim 9, wherein:

It also includes more than one alternate DNS server connected to the DNS server. The data of the DNS server is updated synchronously with the data of the alternate DNS server. When the DNS server fails, the backup DNS server continues to work.