CN100411401C - Method for network equipment self adaption load equalization - Google Patents

Abstract

The present invention relates to a self adaption load equalization method for a network device. Each node in a network device cluster makes use of autonomous load equalization operation to process network data packets which enter the cluster in parallel and in the equalization mode. Then, load information, a, CPU and memory utilization ratio information are obtained from each node. When failure of the operating state of a certain node occurs, the cluster can dynamically adjust the rest of normal nodes, and can ensure that the cluster can normally process all network data packets. The present invention solves the problem of the load equalization of network flow quantity, improves the reliability, the high availability and the processing capability to the data packets of the network system, gives full play to the processing capability and the utilization ratio of a plurality of network devices, and solves the problems of the single failure point of the system.

Description

Self-adaptive load balancing method for network equipment

The technical field is as follows:

the invention relates to a network equipment self-adaptive load balancing method, which comprises a firewall, a router, a Virtual Private Network (VPN) gateway, intrusion detection equipment or a special server and other load balancing methods; in particular to a method for realizing distributed self-adaptive load balance of a network cluster, belonging to the technical field of computer networks.

Background art:

at present, the exponential increase of network traffic makes network devices such as firewalls, routers and the like become bottlenecks that limit network bandwidth, and greatly restricts the practical application of networks. Meanwhile, the reliability and high availability of the whole network system are greatly influenced. In order to improve the reliability of the network system, a common technical means is to adopt a firewall dual-computer hot standby technology. However, in the firewall dual-computer hot standby technology, the firewall as the hot standby is in the standby state and does not participate in processing the network traffic. The essence of this technique is to ensure the reliability of the system through redundant techniques. Although this improves the reliability of the entire system to some extent, it wastes device resources and cannot fully exploit the utilization and efficiency of firewalls.

On the other hand, network security devices such as firewalls, VPN gateways, etc. increasingly become the single point of failure of network systems. The problem is only relieved by adopting the dual-computer hot standby technology, but the problem cannot be fundamentally solved.

Therefore, advanced technologies must be adopted to solve the load balancing problem of network traffic, so as to improve the reliability, high availability and processing capability of the network system.

The invention content is as follows:

the invention mainly aims to provide a network equipment self-adaptive load balancing method, which is used for solving the load balancing problem of network flow and improving the reliability, high availability and data packet processing capacity of a network system.

Another objective of the present invention is to provide a method for network device adaptive load balancing, which fundamentally solves the problem that a network security device becomes a single failure point of a network system.

The invention is realized by the following technical scheme:

a network equipment self-adapting load balancing method, each node in the network equipment cluster monitors the node running state through heartbeat protocol, each node only processes the data packet belonging to its own processing range, and the load balancing operation in the network equipment cluster is carried out according to the running state of each node; the method specifically comprises the following steps:

step 1: a master node in the cluster communicates with each slave node through a heartbeat protocol to acquire load information and state information of each slave node;

step 2: when the slave node fails, the master node performs load balancing operation in the network equipment cluster again according to the residual node number, and redistributes a data packet hash (hash) interval to be processed for each node.

The above method further comprises:

and step 3: the master node equipment sends heartbeat information of the master node equipment to other slave equipment within a specified time; when the slave device does not receive the heartbeat information of the master node device within the specified time, the master node device is considered to have a fault, and one slave node device near the master node device automatically takes over the task of the master node device.

The heartbeat information at least comprises: node number or load condition or IP address or Hash value space or one or the combination of configuration synchronization mark information or load synchronization mark information; the information is transmitted by each node device to other nodes through the network within a prescribed time.

The load balancing operation in the network equipment cluster comprises the following steps:

step 21: calculating the hash value interval of the whole cluster; the method specifically comprises the following steps: determining the interval of the cluster hash value, setting the lower limit of the interval to be zero, and calculating the upper limit of the hash interval according to the following formula:

HU＝(2^m-1)；

wherein,

HU is the upper limit value of the hash interval;

m is a hash value binary digit value;

step 22: dividing a hash value interval of each sub-device in the cluster according to the following formula;

HashI＝((i-1)×SE，i×SE]

$\mathrm{SE}=\frac{\mathrm{HashV}}{\mathrm{SEN}}$

wherein: 1, 2, …, SEN;

HashI is a Hash numerical value interval where the I-th sub-device is located;

SE is a numerical value interval size value occupied by each sub-device of the cluster;

HashV is the cluster hash upper limit value HU calculated in step 21;

SEN is the total number of sub-devices in the cluster;

step 23: the ports with the same number of all the devices in the cluster can receive network messages from the external interfaces of the cluster through the shared network;

step 24: each node device calculates and processes the received network data message.

The header information includes at least source address and destination address information of the IP header.

The specific processing procedure of step 24 is:

step 241: when each node device receives a network data message, the header information of the message is taken out, and the hash function is used for calculation;

step 242: judging whether the calculation result falls within a hash value interval occupied by the node equipment; if yes, the node equipment processes the received network message; otherwise, the network message is discarded.

Each node device in the network device cluster is the same device, and one of the network interfaces of each node device is connected through a shared network.

The invention solves the load balancing problem of network flow, and improves the reliability, high availability and data packet processing capability of the network system. The processing capacity and the utilization rate of a plurality of network devices are fully exerted, and the problem of single fault point of the system is solved.

Description of the drawings:

fig. 1 is a schematic structural diagram of a network cluster according to the present invention.

The specific implementation mode is as follows:

the invention is illustrated in detail in the following by specific examples and figures:

referring to fig. 1, in an example of the present invention, m node devices are provided in a network cluster 1, and each node device is the same device; the number of network interfaces of each device is 3, and the serial numbers are as follows in sequence: i1, I2, I3, i.e. 3 network interfaces outside the whole network cluster 1.

First, cluster node devices are connected, and network interfaces I1 of m devices are connected by using a shared network device, such as: an Ethernet hub with a broadcasting function or a switch with a broadcasting function, wherein the number of network ports of the hub is at least m + 1; the m ports of the hub are connected to I1 ports of m devices, and the m +1 th interface of the hub is used as the 1 st interface outside the cluster. The same as the method, I2 and I3 interfaces of m devices are connected in sequence to form I2 and I3 network interfaces outside the cluster.

The automatic load balancing method of the network data packet text by the node equipment in the cluster comprises the following steps:

calculating the maximum value of the cluster hash;

in the cluster of this embodiment, there are m sub-devices, then: and equally dividing the cluster hash maximum value range from 0 to the cluster hash maximum value range into m cluster hash value intervals, wherein each device occupies one value interval.

Ports with the same number of all the devices in the cluster can receive network messages from external interfaces of the cluster due to the connection method of the cluster node devices.

When a port of each node device receives a network data message, firstly, the header information of the message is taken out, a hash function is used for calculating a result, then, whether the hash result falls in a hash value interval occupied by the node device is judged, and if yes, the node device processes the received network message; otherwise, the network message is discarded.

By the method, the cluster is processed in parallel, so that the network equipment in the cluster has the capacity of automatic load balancing, the distributed self-adaptive load balancing of the network cluster is realized, and in order to ensure the reliable operation of the cluster, the management of the cluster adopts a master-slave mode management, namely: and setting one node device as a master node and other node devices as slave nodes in the cluster, wherein the cluster nodes perform load balancing according to the load condition of each node, the utilization rate of a CPU (Central processing Unit) and a memory and the load condition of the whole cluster, which are exchanged among the cluster nodes through a heartbeat protocol.

The heartbeat information of the node device described herein includes: node number, load condition, IP address, hash value space, configuration synchronization mark, load synchronization mark and other information.

Each node device sends heartbeat information to the master node device within the specified time, and if the master node device does not receive the heartbeat of the slave node device within the specified time, the slave node device is considered to have a fault; when a certain slave node device fails, the master node device recalculates the cluster hash interval according to the number of the remaining slave node devices, and allocates a new cluster hash interval value to the remaining slave node devices; the master node equipment also has to send own heartbeat information to other slave equipment within a specified time; when the slave node device does not receive the heartbeat of the master node device within the specified time, the master node device is considered to be in failure, and one slave node device near the master node device automatically takes the task of the master node device.

Finally, it should be noted that: the above embodiments are only used to illustrate the present invention and do not limit the technical solutions described in the present invention; thus, while the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted; all such modifications and variations are intended to be included herein within the scope of this disclosure and the present invention and protected by the following claims.

Claims (7)

1. A method for network device adaptive load balancing is characterized in that: each node in the network equipment cluster monitors the node running state mutually through a heartbeat protocol, each node only processes data packets belonging to the processing range of the node, and load balancing operation in the network equipment cluster is carried out according to the running state of each node; the method specifically comprises the following steps:

step 1: a master node in the cluster communicates with each slave node through a heartbeat protocol to acquire load information and state information of each slave node;

step 2: and when the slave node fails, the master node performs load balancing operation in the network equipment cluster again according to the residual node number, and redistributes the hash interval of the data packet to be processed for each node.

2. The method of network device adaptive load balancing according to claim 1, wherein: the method still further comprises:

and step 3: the master node equipment sends heartbeat information of the master node equipment to other slave equipment within a specified time; when the slave device does not receive the heartbeat information of the master node device within the specified time, the master node device is considered to have a fault, and one slave node device near the master node device automatically takes over the task of the master node device.

3. The method for network device adaptive load balancing according to claim 1 or 2, wherein: the heartbeat information at least comprises: node number or load condition or IP address or Hash value space or one or the combination of configuration synchronization mark information or load synchronization mark information; the information is transmitted by each node device to other nodes through the network within a prescribed time.

4. The method of network device adaptive load balancing according to claim 1, wherein: the load balancing operation in the network equipment cluster comprises the following steps:

step 21: calculating a hash interval of the whole cluster; the method specifically comprises the following steps: determining the interval of the cluster hash value, setting the lower limit of the interval to be zero, and calculating the upper limit of the hash interval according to the following formula:

HU＝(2^m-1)；

wherein,

HU is the upper limit value of the hash interval;

m is a hash value binary digit value;

step 22: dividing a hash value interval of each sub-device in the cluster according to the following formula;

HashI＝((i-1)×SE，i×SE]

$\mathrm{SE}=\frac{\mathrm{HashV}}{\mathrm{SEN}}$

wherein: 1, 2, …, SEN;

HashI is a Hash numerical value interval where the I-th sub-device is located;

SE is a numerical value interval size value occupied by each sub-device of the cluster;

HashV is the cluster hash upper limit value HU calculated in step 21;

SEN is the total number of sub-devices in the cluster;

step 23: the ports with the same number of all the devices in the cluster can receive network messages from the external interfaces of the cluster through the shared network;

step 24: each node device calculates and processes the received network data message.

5. The method of network device adaptive load balancing according to claim 4, wherein: the header information includes at least source address and destination address information of the IP header.

6. The method of network device adaptive load balancing according to claim 4, wherein: the specific processing procedure of step 24 is:

step 241: when each node device receives a network data message, the header information of the message is taken out, and the hash function is used for calculation;

step 242: judging whether the calculation result falls within a hash value interval occupied by the node equipment; if yes, the node equipment processes the received network message; otherwise, the network message is discarded.

7. The method for network device adaptive load balancing according to claim 1, 2, 4, 5 or 6, wherein: each node device in the network device cluster is the same device, and one of the network interfaces of each node device is connected through a shared network.

Images