KR101014977B1 - Load balancing method in the function of link aggregration - Google Patents

Abstract

The present invention relates to a load balancing method in a link integration function. In a switching system using a link integration function, a load balancing module is called if there are two or more member ports in a link integration ID periodically through a main module. Making a step; Determining whether to modify a hash map required to perform load balancing by selecting ports whose traffic amount corresponds to a predetermined reference value through the called load balancing module, and calling a hash map modification module if modification is necessary; And modifying the hash map for the member port selected through the called hash map modification module.

According to the present invention, the link aggregation function can prevent traffic from flowing to a specific member port and prevent the loss of packets.

Load balancing, main module, load balancing module, hash map modification module, link integration

Description

Load balancing method in link aggregation function {LOAD BALANCING METHOD IN THE FUNCTION OF LINK AGGREGRATION}

1 is an exemplary flowchart illustrating a link integration function setting process according to the prior art;

2 is an exemplary flowchart illustrating an example of determining an output port of a packet during link aggregation function setting according to the prior art;

3 is an exemplary block diagram showing an example in which traffic is heavy when setting a link aggregation function according to the prior art;

4 is a flowchart of a main module for explaining the present invention;

5 is a flowchart of a load balancing module for explaining the present invention;

6 is a flowchart of a hash map modification module for explaining the present invention;

The present invention relates to a load balancing method in a link aggregation function that prevents traffic from being gathered to a specific member port and prevents packet loss by enabling proper load balancing in a switching system using a link integration function.

In general, load balancing is the balancing of the load on each processor in a multiprocessor system in which multiple processors process work in parallel. It means distributing work properly so that it is not wasted and, if necessary, moving work from one processor to another.

In other words, different types of traffic are properly distributed to different ports.

In addition, link aggregation refers to a method of organizing multiple physical links into a single logical link. For example, two 100 Mbps Ethernet links can be bundled together and used as one Ethernet link. You can think as you do.

In other words, link aggregation is used when one physical link cannot meet the bandwidth required by the user.

However, when 200Mbps is set using the link aggregation function, the speed of 200Mbps is not exactly obtained because data traffic is not divided by exactly 1/2 on two physical links.

This is because link aggregation uses a hash function to set the probability to 1/2, where hash is a value that determines the destination using a specific operation (such as the total sum). Way.

Conventionally, when the link integration function is set, a part for creating an ID and a hash map, a part for making a hash key using packet information of a packet to be output, and a hash map with a hash map).

That is, as shown in FIG. 1, when the link aggregation function is activated, all physical links set to link aggregation have one unified ID (S10 and S11).

Subsequently, a hash map is set such that all physical links are equally probabilistic (S12).

For example, suppose you have a 16-bit hash map, and if four physical links are link aggregated, you can use hashes such as F000 (port 1), 0F00 (port 2), 00F0 (port 3), and 000F (port 4) in hexadecimal. A map can be specified.

On the other hand, when mapping a hash map with such a hash, as shown in FIG. 2, if there is data to be outputted, and if there is data to be outputted, a hash with link integration is generated by generating a hash with information of packet data to be outputted. After checking the maps, this key is used to transmit data to either port of the integrated link (S20 to S23).

Here, the information used to make the hash may be a Mac address, an IP address, or a Layer 4 address, and all of them use the information of the packet to be output.

This is an important factor in deciding which port to send to. However, depending on the type of packet, hashing can be weighted to a specific port of the hash map. For this reason, link aggregation is often referred to as load sharing instead of load balancing.

For example, as shown in FIG. 3, when three types of packets (Packets A, B, and C) should be output, the calculated packet is 8000 (hexadecimal) and Packet B is 800 (16). If packet C is 2, all three packets are output to port 1.

At this time, if port 1 and port 2 are bundled with link aggregation function, and port 1 is a hash map of AAAA (hexadecimal) and port 2 is 5555 (hexadecimal), such link integration alone can guarantee load balancing. There will be no.

That is, when the link aggregation function is used, traffic may be crowded to a specific member port.

The present invention was created in view of the above-described problems of the prior art, and solves this problem. Periodically, the traffic is periodically checked to check whether the member ports integrated with the link are load-balanced or at the highest traffic. Load balancing in link integration function that can improve load balancing when using link aggregation function by moving hash map to least member port or checking traffic several times when moving hash map and preparing for abnormal traffic. Its main purpose is to provide a method.

In order to achieve the above technical problem, in the switching system using the link integration function, calling the load balancing module if it is periodically checked whether there are two or more member ports in the link integration ID through the main module. Wow; Determining whether or not to modify the hash map necessary to perform load balancing by selecting the ports of the member port traffic amount of the predetermined value through the called load balancing module, and calling a hash map modification module if necessary ; The method provides a load balancing method in a link integration function including the step of modifying a hash map for the member port selected through the called hash map modification module.

In addition, the present invention is characterized in that to modify the hash map to move the hash map from the member port with the most traffic to the member port with the least traffic in the hash map modification step.

In addition, the present invention is characterized in that the traffic is checked several times during the movement of the hash map so that no transient abnormality is generated.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment according to the present invention.

4 is a flow chart showing a processing flow in the main module for explaining the present invention, Figure 5 is a flow chart showing a processing flow in the load balancing module for explaining the present invention, Figure 6 illustrates the present invention This is a flowchart showing the processing flow in the hash map modification module.

The present invention consists of a main module, a load balancing module and a hash map modification module.

At this time, the main module periodically performs a function of calling a load balancing module for an ID having a member port of 2 or more by finding a link integration ID.

Then, the load balancing module finds the member port (maxport) with the most traffic and the member port (minport) with the least traffic, determines whether the hash map modification is necessary, and performs a function of calling a hash map modification module if necessary. to be.

The hash map modification module is a module that actually modifies the hash map, and checks whether load balancing is performed by moving one bit of maxport's hashmap to minport.

These load balancing modules and hash map modification modules are called during processing of the main module to distribute the functions of only necessary parts.

Then, a load balancing method according to the present invention will be described.

First, as shown in FIG. 4, an initial inspection time, that is, time_interval is set (S100).

time_interval is a value set by the user at intervals during which load balancing is performed, and the smaller the value is, the faster the traffic change can be made. On the other hand, the larger the value, the smaller the change in the previously set hashmap. There is a negligible effect on instantaneous traffic fluctuations.

At this time, it is particularly preferable to periodically check traffic through time_interval.

Subsequently, the member port is checked for all the link integration IDs and the load balancing module is called for the IDs having two or more member ports (S110, S120, and S130).

When a specific link integration ID is called from the main module through the step S130, as shown in FIG. 5, each traffic amount is checked for the member port belonging to this ID (S200).

As a result of the inspection, the member port (maxport) with the most traffic amount and the member port (minport) with the least traffic amount are selected among them (S210).

Subsequently, it is checked whether the amount of traffic of Maxport is greater than thresh1. If the amount of traffic of Maxport is greater than thresh1, then the amount of traffic of Minport is then less than thresh2 (S220, S240).

At this time, thresh1 may be a specific traffic amount that the user wants to change the hash map, or may select the moment when the packet is dropped (when the amount of packets to be output is greater than the maximum amount of the corresponding port).

In addition, if the traffic amount of the maxport (maxport) is less than thresh1 in step S220, since the hash map does not need to be changed, the variable is initialized and then returned to the main module (S230).

In addition, the thresh2 value is to prevent load balancing when the traffic of the minport to distribute the traffic is too large. The meaning of the traffic of the minport is that the traffic already exists in the ports in the link aggregation. It means it's full.

Then, if the traffic amount of the minport (minport) is greater than thresh2 in step S240, since the hash map does not need to be changed, the variable is initialized and then returned to the main module (S250).

On the other hand, if both of the steps S220 and S240 is satisfied, the two conditions are checked by thresh3 (S260).

In this case, the reason for placing thresh3 is to prevent the hash map from being changed by a momentary traffic change.

Of course, this value can be set to 1 to immediately adapt to instantaneous traffic changes.

If the three conditions are satisfied, the hash map modification module having the processing flow as shown in FIG. 6 is called.

At this time, the thresh1 refers to the minimum amount of traffic required to remove the bit of the hash map, thresh2 refers to the maximum amount of traffic that can be added bit of the hash map, thresh3 refers to the number of re-check to ignore the instantaneous traffic.

When the hash map modification module is called through the above-described process, first, it is checked whether the corresponding bit is 1 in each bit of the Maxport hash map (S300 and S310).

Here, the hash map modification module is actually a module that modifies the hash map.

As a result of the check, if the corresponding bit is 1, the bitmap is subtracted from the maxport and added to the hashmap of the minport, and then the hashmap is modified (S320 and S330).

That is, since traffic corresponding to 1 bit of Maxport is moved to Minport, the hash map needs to be modified.

When the hash map is modified, the traffic volume of Maxport and Minport is checked to check whether both are smaller than thresh1 (S340 and S350).

If the result of the check is small, the load balancing is correct, and the main module is returned. If not, the process of restoring the modified hashmap and restoring the other hashmap bits of Maxport is repeated. The process returns to step S310 (S360, S370).

On the other hand, if the bit is not 1 in step S310 immediately returns to the main module to perform the original load balancing.

As a result, when the link aggregation function is used, traffic can be prevented from flowing to a specific member port.

As described in detail above, according to the present invention, when the link aggregation function is used, traffic is prevented from being directed to a specific member port, which causes traffic to be directed to a specific member port despite being able to accommodate more traffic on the line. This can prevent packets from being lost.

Claims (3)

In switching systems that use link aggregation, Invoking the load balancing module if there are two or more member ports in the link integration ID periodically through the main module; Determining whether to modify a hash map required to perform load balancing by selecting ports whose traffic amount corresponds to a predetermined reference value through the called load balancing module, and calling a hash map modification module if modification is necessary; And modifying a hash map for the member port selected through the called hash map modification module. The method according to claim 1; The hash map modification step, A load balancing method in a link aggregation function, which moves a hash map from a member port having the most traffic to a member port having the least traffic. The method of claim 2, wherein the hash map movement, A load balancing method in a link aggregation function, characterized in that the traffic is checked several times to prevent instantaneous abnormal traffic.

Images