So, every article I've read and Youtube video said that Minimal APIs are way faster than the "old" Web APIs. Since I already had this project setup, I decided to put that to the test.
The set up was easy, because I already had the adapters (AppServices), which will guarantee me that no API will have unfair advantages. Also in the name of fairness, each API was using a different database file.
In this scenario, for both APIs the target of the test was a get by id route, returning the same record over and over again.
test info
test suite: nbomber_default_test_suite_name
test name: nbomber_default_test_name
session id: 2023-03-01_02.02.33_session_b237a78d
scenario stats
scenario: Minimal API
-
ok count:
28800 -
fail count:
0 -
all data:
11,7MB -
duration:
00:01:00
load simulations:
inject, rate:24, interval:00:00:00.0500000, during:00:01:00
| step | ok stats |
|---|---|
| name | global information |
| request count | all = 28800, ok = 28800, RPS = 480 |
| latency | min = 0,19, mean = 0,88, max = 31,46, StdDev = 0,7 |
| latency percentile | p50 = 0,79, p75 = 1,03, p95 = 1,39, p99 = 1,75 |
| data transfer | min = 0,417 KB, mean = 0,417 KB, max = 0,417 KB, all = 11,7 MB |
status codes for scenario:
Minimal API
| status code | count | message |
|---|---|---|
| OK | 28800 |
scenario stats
scenario: Web API
-
ok count:
28800 -
fail count:
0 -
all data:
12,3MB -
duration:
00:01:00
load simulations:
inject, rate:24, interval:00:00:00.0500000, during:00:01:00
| step | ok stats |
|---|---|
| name | global information |
| request count | all = 28800, ok = 28800, RPS = 480 |
| latency | min = 0,39, mean = 1, max = 32,29, StdDev = 0,46 |
| latency percentile | p50 = 0,95, p75 = 1,16, p95 = 1,51, p99 = 1,84 |
| data transfer | min = 0,438 KB, mean = 0,438 KB, max = 0,438 KB, all = 12,3 MB |
status codes for scenario:
Web API
| status code | count | message |
|---|---|---|
| OK | 28800 |
Although Minimal APIs were faster, it wasn't by much. The difference was only 0.12 seconds, which is not a lot. Both were able to handle 480 requests per second during 1 minute.
To test the performance of the API, this tests uses only and endpoint that returns the current datetime. No services, no dependency injection, nothing.
test info
test suite: nbomber_default_test_suite_name
test name: nbomber_default_test_name
session id: 2023-03-01_02.53.62_session_501e17e1
scenario stats
scenario: Minimal API
-
ok count:
28800 -
fail count:
0 -
all data:
2,6MB -
duration:
00:01:00
load simulations:
inject, rate:24, interval:00:00:00.0500000, during:00:01:00
| step | ok stats |
|---|---|
| name | global information |
| request count | all = 28800, ok = 28800, RPS = 480 |
| latency | min = 0,19, mean = 0,79, max = 15,5, StdDev = 0,46 |
| latency percentile | p50 = 0,74, p75 = 0,94, p95 = 1,28, p99 = 1,56 |
| data transfer | min = 0,088 KB, mean = 0,091 KB, max = 0,092 KB, all = 2,6 MB |
status codes for scenario:
Minimal API
| status code | count | message |
|---|---|---|
| OK | 28800 |
scenario stats
scenario: Web API
-
ok count:
28800 -
fail count:
0 -
all data:
2,6MB -
duration:
00:01:00
load simulations:
inject, rate:24, interval:00:00:00.0500000, during:00:01:00
| step | ok stats |
|---|---|
| name | global information |
| request count | all = 28800, ok = 28800, RPS = 480 |
| latency | min = 0,26, mean = 1,07, max = 32,4, StdDev = 0,72 |
| latency percentile | p50 = 1,03, p75 = 1,22, p95 = 1,53, p99 = 1,8 |
| data transfer | min = 0,088 KB, mean = 0,091 KB, max = 0,092 KB, all = 2,6 MB |
status codes for scenario:
Web API
| status code | count | message |
|---|---|---|
| OK | 28800 |
This time, the difference was bigger. Minimal APIs were 0.28 seconds faster than Web APIs. Maybe the database was the bottleneck in the previous test. Still, not a big difference.
I might try change the database, maybe use an in-memory database. It was my first time trying out NBomber, so there's the distinct possibility of me doing something wrong. I might also try to add more endpoints to the Web API, to see if it can handle more requests.