Not as you may think: the non-equivalent Server Load balancer of the network type.

Remember that there is a question in NP about the non-equivalent Server Load balancer of China. The answer is controversial. Today, I'm going to verify the non-equivalent Server Load balancer of the network.

First of all, the non-equivalent Server Load balancer of VPN is a traffic load formed between Successor and Feasible Successor. This type of Server Load balancer is unique to VPN. However, the Kerberos ququal Load Balancing must meet two conditions (maximum does not need to be considered. The default value is four paths ):

1. Feasible Successor must exist (FS is used below)

Feasible Successor must meet the following requirements: AD of secondary-best route <FD of best route (Successor ).

There is no dispute over FS.

2. Configure the Variance value

FD of FS route <FD of best route (Successor) * Varince

In this regard, some people think it is fs fd.Equal to or lessSuccessor * varince, with instructions on the cisco Official Website:

Http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a008009437d.shtml.

This configuration increases the minimum metric to 40 (2*20 = 40). mongodes all routes that have a metric
Less than or equal to 40 and satisfy the feasibility condition.

Let's verify it now.

Top:

 

Note:By default, the calculation of the MongoDB metric is (10 to the power of 7/bandwidth (kpbs) + delay/10) * 256. The default value is 2 k vaule for bandwidth and latency. Considering metric calculation, latency is used as the unique metric parameter.

Purpose:

Non-equivalent Server Load balancer

Experiment process:

On R1, follow the route entries that reach the destination CIDR Block 45.1.1.0/24.

Change the latency of interfaces serial0/0 and serial0/1 of R1 to 10 and 40, respectively, the delay between the serial0/1 interface of R2 and the serial0/0 interface of R3 and the serial0/2 interface of R4 is changed to 10.

Note:Latency is the latency value of the accumulated ingress interface.

1. Do not modify the variance value first. The configuration of the OSPF above R1 is as follows:

 

2. view the IP protocol

Note:When establishing a neighbor relationship, the necessary conditions are: AS number, K value, consistent authentication, and need to be a direct connection network segment.

3. The neighbor table of the OSPF protocol

4. Topology table

Here we can see that 12.1.1.2 is the successor that reaches 45.1.1.0/24, and 13.1.1.3 is FS. WhileFs fd is exactly twice that of successor FD. Because the variance value is not configured

R1 route table:

Only the 45.1.1.0/24 CIDR block has one path.

The variance value is now set to 2 in the r1's OSPF process.

Note that: the update is triggered by the replication of the network. After the variance is modified, only the topology changes. Therefore, we need to clear the next neighbor relationship: clear ip: VPN. net.

Ip protocol of R1 after re-convergence (maximum can be modified without modification ):

R1 topology table (variance = 2)

The topology table does not change. Note: The FD of FS reaches 45.1.1.0/24 is twice that of successor FD. The result is:

R1 route table:

Test results:

R1 reaches 45.1.1.0/24 for non-equivalent load balancing.

FD of FS route <FD of best route (Successor) * Varince (this is inaccurate)

 

Real results:

FD of FS route <= FD of best route (Successor) * Varince

From Chao Hua Xi qiboCustomer