Understanding JMeter results on simple Laravel application - laravel

Question

I am struggling to make sense of the load test results from Jmeter.
I want to understand how much load a plain vanilla Laravel application can handle. I setup simple endpoints on plain Laravel 8 and tried on different AWS EC2 instances (t3.medium, t3.xlarge, load balanced setup etc).
The following routes were used were we forcefully increase the response time:
Route::get('/r0', function (Request $request) {
$data = [
'name' => 'test_name',
'image' => 'imageURL',
'translations' => [
'en' => 'test_english',
'te' => 'పంచాయతి సెక్రటరి',
],
];
return
response()->json($data, 200, ['Content-Type' => 'application/json;charset=UTF-8', 'Charset' => 'utf-8'], JSON_UNESCAPED_UNICODE);
});
Route::get('/r1', function (Request $request) {
sleep(2);
$data = [
'name' => 'test_name',
'image' => 'imageURL',
'translations' => [
'en' => 'test_english',
'te' => 'పంచాయతి సెక్రటరి',
],
];
return
response()->json($data, 200, ['Content-Type' => 'application/json;charset=UTF-8', 'Charset' => 'utf-8'], JSON_UNESCAPED_UNICODE);
});
Route::get('/r2', function (Request $request) {
sleep(10);
$data = [
'name' => 'test_name',
'image' => 'imageURL',
'translations' => [
'en' => 'test_english',
'te' => 'పంచాయతి సెక్రటరి',
],
];
return
response()->json($data, 200, ['Content-Type' => 'application/json;charset=UTF-8', 'Charset' => 'utf-8'], JSON_UNESCAPED_UNICODE);
});
I recorded the throughput and deviation as obtained from the “Graph results” and other details as obtained from the “Aggregate Report” on Jmeter for different end points.
Here’s a sample result for r1 on one of the setups:
From the above image/table, here are some of the questions I want answers for:
What is the load that server can handle for this end point? Is it the point where deviation is higher than throughput (as said in some of the blog posts on the internet) or when the error % is non zero?
How does one define the max load or capacity the server can handle? Can it be an absolute number like 200 users at any point?
For some end points, the server stops completely responding after a certain load. I had to restart the server before I could do further testing. Why does that happen?

Answer 1

Your table doesn't tell the full story regarding how many threads were active and what was the relationship between the number of threads, throughput, response time, errors, etc. I would rather suggest generating HTML Reporting Dashboard, it's way more informative than the aggregated numbers
Normally I would look for the saturation point - the point of maximum system performance, like:
start with 1 virtual user
gradually increase the load observing i.e. Transactions per Second chart. On well behaved system the throughput (number of transactions per second) should increase by the same factor as the number of virtual users, i.e. you increase the load twice - throughput should be twice higher. response times should be more or less equal
at some point you will see that throughput decreases and response time increases. if you look at Active Threads Over Time chart right before this moment - you will see how many virtual users were online at this stage - this is the number you're looking for
you can continue increasing the load to see when the errors start occurring or application terminates
I think point one provides the answer
There are many possible reasons, the most common/obvious are:
the server lacks essential resources like CPU, RAM, Network, it can be checked using JMeter PerfMon Plugin
the server is not properly configured for high loads, see i.e. 12 Tips for Laravel Performance Optimization in 2020 guide for example tuning tweaks
check your application and operating system logs, it might be the case it has been terminated due to consuming previously mentioned resources for example by OOM Killer