Breaking the 300 barrier

1

Puma

As we use Puma as our webserver for our rails application, I quickly went to Puma's config file which typically resides in config/puma.rb. The config was set as.

#Web And Application Servers #Application Server #Web Servers 3 social mentions

2

NewRelic

Recently, we have been experiencing some Latency issues. The reason being the increase in our userbase as more companies are adopting our platform to do their procurement efficiently. Our core product is a monolith rails application served on AWS servers. We use AWS ELB's Application load balancer to swiftly balance the load on multiple servers and increase them when needed. So, these sudden latency issues caused Load Balancer to aggressively scale up horizontally by adding more servers. This worked and made our latency back to an acceptable 200ms limit. But the catch is that this approach is not reliable enough to keep up with the cost. After viewing the data from New Relics, we found out that our one server can go upto at max of 300 RPM limit and we thought this is what max a server can achieve.

#Application Performance Monitoring #Performance Monitoring #Monitoring Tools 80 social mentions

3

locust

So, in theory, we can handle 300 requests per minute on a single server which was the assumption we started with. After this, I decided to play with this configuration and see what we could achieve. But, to go ahead I need a system to measure the metrics of our load testing. So I quickly set up Locust on my system. Locust is an open-source easy to setup load-testing framework.

#Analytics #Web Analytics #Privacy 55 social mentions