I currently am using mod_proxy_balance in Apache for load balancing. I have a client emulator which creates 100 threads and emulates user activities like browse the website. I tried making the client talk to the webserver directly and also through the load balancer. Here is what I found.
When client talks directly to webserver:
- delay = 28ms
When client talks to webserver through a loadbalancer (only one backend webserver):
- delay = 53ms
When client talks to webserver through a loadbalancer with two backend webservers:
- delay = 113ms
The delay keeps increasing, is there a way to get this delay down? It is a very important aspect to provide agreeable SLO.