0
I created a Load Balancer. When testing internally in the Load Balancer Server using wrk
, I received a benchmark of 300,000 Requests Per Second.
. Then, I created a droplet in Digital Ocean and ran the same wrk test, this time externally from digital ocean to the internal load balancer server, with results of 5,000 Request per Second
.
Does the low result have anything to do with my server resources and/or code, or is it completely due to the resources on the droplet.?
If someone could elaborate on the situation and what they believe is the cause of the low requests, id be appreciative. Maybe nothing is wrong at all with the load balancer i created and thats expected as the internal wrk test represents what the load balancer server is capable of whereas the external results is based on their internal hardware and can only produce as best as they can?
Any information in any way would be helpful.
Clearly there is a difference between testing while directly connected to the load balancer and testing over the internet from a remote machine. There are several factors to consider: bandwidth, testing machine resources, distance and latency, etc. But it is unreasonable to think you’re going to get anywhere near similar performance testing over the internet compared to locally connected. – Appleoddity – 2018-10-17T04:22:06.170
@Appleoddity But such a dramatic decrease? Going from 300K to 5K is astronomical. Does this seem normal then, around 5K – Jonathan Kushner – 2018-10-17T04:24:18.953
I honestly can’t tell you what is “normal.” There are far too many factors and I know nothing about your load test configuration, your load balancer config, your app servers or web servers, etc. This is not a question I can answer. But yes, there will be a dramatic decrease in performance when tested remotely as opposed to locally and yes if the remote instance is underpowered it could play a significant role in how digital ocean is allocating bandwidth and other resources to your instance. – Appleoddity – 2018-10-17T04:29:23.893