Name

ngx_http_upstream_vnswrr_module.

Introduction

The VNSWRR module is an efficient load balancing algorithm that is smooth, decentralized, and high-performance compared to Nginx's official SWRR algorithm.

Example

http {

    upstream backend {
        vnswrr; # enable VNSWRR load balancing algorithm.
        127.0.0.1 port=81;
        127.0.0.1 port=82 weight=2;
        127.0.0.1 port=83;
        127.0.0.1 port=84 backup;
        127.0.0.1 port=85 down;
    }
    
    server {
        server_name localhost;
        
        location / {
            proxy_pass http://backend;
        }
    }
}

Installation

Build Tengine with this module from source:


./configure --add-module=./modules/ngx_http_upstream_vnswrr_module/
make
make install

Directive

vnswrr

Syntax: vnswrr [max_init=number]
Default: none
Context: upstream

Enable vnswrr load balancing algorithm.

max_init

The max number of virtual node per initialization, to avoid initializing too many virtual nodes in one request.

In very large clusters, setting this argument can significantly reduce the time overhead of a single request.

There are two additional rules of max_init:
1. If max_init is not set or 0, it will be adjusted to the number of peers.
2. If max_init is greater than total weight of peers, it will be adjusted to the total weight of peers.

http {

    upstream backend {
        # at most 3 virtual node per initialization
        vnswrr max_init=3;
        127.0.0.1 port=81 weight=101;
        127.0.0.1 port=82 weight=102;
        127.0.0.1 port=83 weight=103;
        127.0.0.1 port=84 weight=104;
        127.0.0.1 port=85 weight=105;
    }
    
    server {
        server_name localhost;
        
        location / {
            proxy_pass http://backend;
        }
    }
}

Performance

Under the same pressure (wrk, 500 concurrency, keepalive, 2000 endpoint), the CPU consumption of VNSWRR algorithm accounts for 0.27% ( ngx_http_upstream_get_vnswrr). Compared with VNSWRR algorithm, the CPU consumption of SWRR (ngx_http_upstream_get_peer 39%) is an order of magnitude higher than VNSWRR.

In the above environment, the QPS of VNSWRR increased by 60% compared with SWRR algorithm.

Observing the changes of QPS and RT in the different back-end number scenarios. Under SWRR algorithm, with every 500 back-end servers added, the QPS of Nginx decreases by about 10% and RT increases by about 1 ms. But under the VNSWRR algorithm, QPS and RT do not change much.

Ngx Http Upstream Vnswrr Module