Shulou(Shulou.com)06/02 Report--

ARR load balancing based on IIS

Load balancing based on NLB

These two articles are divided into ARR and NLB basic configuration, respectively. Let's talk about how to combine them to form a highly available / scalable cluster.

What is highly available and scalable?

High availability: means that the system is specially designed to reduce downtime and keep services highly available. Normal working hours / unit time (usually one year), the ratio is more than 99.99%, that is, one hour of shutdown. It can be considered as high availability.

Scalable: refers to the difficulty of increasing or decreasing the hardware to increase or decrease the performance of the software.

Don't worry about the ARR cluster, because it has its own health check, and if there is an exception on the sub-server, the reverse proxy will not convert the traffic to the corresponding server.

Now I am more worried about a problem. What if an exception occurs on the proxy server? The proxy server is responsible for the daily processing of conversion requests and does not rule out the possibility of failure. ARR servers are prone to a single point of failure, so the entire ARR deployment does not have high availability.

The number of sub-servers is actually limited by the forwarding ability of the ARR server, that is, scalability is limited.

How to solve ARR single point of failure to achieve high availability?

First, let's build two ARR servers.

192.168.99.8 192.168.99.10

Install NLB on these two servers, together with NLB load balancer.

Note that the host priority is currently set to 192.168.99.8 as the priority server.

The cluster port rule sets port 80.

Set to a single host

In other words, of the two ARR, only one is working, that is, 192.168.99.8, and the other 192.168.99.10 is a failover server. Deployment complete.

Some people ask, will there be conflicts in the application if two are deployed?

Actually, no, ARR is on the seventh layer of the application, and NLB is applied on the third layer, the IP layer. In other words, after NLB processing, it will be transferred to ARR for processing.

Overall architecture diagram

Damn it, I turned off visio before I finished drawing. The ARR server on the right should be moved to the right, and the line corresponds to the line. You look old and upset.

Display effect

A total of 150 requests per second.

The total number of requests for the three content servers is 150 per second. It matches.

ARR workload, 192.168.99.8 normal operation, 192.168.99.10 standby. Match

192.168.99.8 requests 150 per second. It matches.

Let's test the failover and suddenly turn off 192.168.99.8. About half a minute later, the request is transferred to 192.168.99.10. We can clearly see that nearly 3000 requests failed.

The test achieved the desired results and coincided.

How to solve ARR scalability?

The above example shows how to build high availability. But the content server is still limited by the conversion ability of the ARR server. How to break through the limited conversion ability of one server to multiple ARR servers to handle this transformation together, we can achieve the scalability of the content server.

The answer is very simple. We just need to change the NLB server cluster, port rules to

Just activate two ARR servers at the same time, the bearing capacity will be doubled, and more content servers can be expanded.

For security reasons, the number of two ARR should be increased to three. If there is a downtime, the traffic will be transferred to ensure that the remaining two ARR servers are within their capacity. Random servers increase, and so on.

The above scheme has been implemented, highly available and scalable. That's all for the notes.

This windows NLB+ARR implementation of Web load balancing highly available / scalable method is the editor to share with you all the content, I hope to give you a reference, but also hope to support more.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.