Introduction to Server Load balancer in RAC

I once talked to my colleagues about the Server Load balancer in RAC. When talking about the distribution mechanism on the server side, we couldn't tell you clearly. I would like to summarize it for future use.

I once talked to my colleagues about the Server Load balancer in RAC. When talking about the distribution mechanism on the server side, we couldn't tell you clearly. I would like to summarize it for future use.

I once talked to my colleagues about the Server Load balancer in RAC. When talking about the distribution mechanism on the server side, we couldn't tell you clearly. I would like to summarize it for future use.

Load Balance

Client load balance

Server load balance

Client load balance

When the client initiates an application request, it first selects a random vip address from tnsnames. ora and then initiates the request through this vip address. This vip address is actually the core of connection high availability. When the instance accessed by the client goes down, and even the vip has been moved to other surviving nodes, the connection will fail. The notification service (FAN) immediately notifies the client to reconnect with other VIPs. The client selects other VIPs in tnsnames. ora to continue the attempt until the connection is successful. Can I use an IP address to connect directly without using the vip connection? The answer is yes, but it takes 30 seconds to notify the client after the connection fails. The price of 30 seconds is too high, and the meaning of high availability is completely lost.

The load balance of the client does not judge the load of the current server. It only distributes the load randomly. Scanning, which includes 11gR2, does not determine the server load. It is distributed round-robin based on the three registered vip addresses. How is the server load determined? Let's continue to look at it.

Server load balance

Server Load balancer works with the remote listener and local listener listeners. The listener registers the current Load information in the LBA (Load Balance Advisor) and can set the Load type based on the service time and throughput. For Web services, you can configure service time as the condition for Server Load balancer, and for query-intensive operations, you can configure the throughput type.

To sum up the above description, the client achieves connection high availability through vip, and the server implements load balance through remote listener, local listener, and LBA configuration.

There are two concepts in load balance:

FAN: Fast Application Notification

Fast notification and proactive push mechanism. It notifies both nodes between clusters and clients to clusters.

LBA: Load Balance Advisor

Server Load balancer recommender

This article permanently updates the link address: