This is a demonstration of a Ratpack application that leverages Spring Boot and makes an asynchronous database call to PostgreSQL. This is from the Spring One Platform 2016 talk, "High Performance Microservices with Ratpack and Spring Boot". This talk demonstrates leveraging Ratpack as the high throughput, non-blocking web layer to an application that is built with the comforts and familiarities of Spring Boot.
With no additional server configuration or tuning, the following results were recorded:
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| t2.micro | t2.micro | 9,529 |
# ./wrk -t4 -c72 -d60s -R10000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 171.367ms, rate sampling interval: 930ms
Thread calibration: mean lat.: 199.882ms, rate sampling interval: 1000ms
Thread calibration: mean lat.: 204.732ms, rate sampling interval: 1061ms
Thread calibration: mean lat.: 187.197ms, rate sampling interval: 983ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.45s 1.13s 4.34s 59.59%
Req/Sec 2.39k 102.57 2.52k 82.91%
571768 requests in 1.00m, 190.30MB read
Requests/sec: 9529.38
Transfer/sec: 3.17MB
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| t2.medium | t2.micro | 10,259 |
# ./wrk -t4 -c72 -d60s -R15000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 1557.261ms, rate sampling interval: 5705ms
Thread calibration: mean lat.: 1553.418ms, rate sampling interval: 5693ms
Thread calibration: mean lat.: 1555.895ms, rate sampling interval: 5701ms
Thread calibration: mean lat.: 1557.657ms, rate sampling interval: 5709ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 11.04s 4.57s 19.22s 57.52%
Req/Sec 2.56k 22.48 2.60k 75.00%
615528 requests in 1.00m, 204.87MB read
Requests/sec: 10258.71
Transfer/sec: 3.41MB
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| t2.medium | t2.medium | 17,399 |
# ./wrk -t4 -c72 -d60s -R20000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 668.483ms, rate sampling interval: 2269ms
Thread calibration: mean lat.: 666.548ms, rate sampling interval: 2248ms
Thread calibration: mean lat.: 677.119ms, rate sampling interval: 2285ms
Thread calibration: mean lat.: 669.698ms, rate sampling interval: 2271ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 4.33s 1.84s 8.08s 57.61%
Req/Sec 4.36k 120.14 4.52k 67.06%
1043910 requests in 1.00m, 347.45MB read
Requests/sec: 17398.57
Transfer/sec: 5.79MB
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| m4.large | t2.medium | 13,884 |
# ./wrk -t4 -c72 -d60s -R20000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 1405.393ms, rate sampling interval: 4927ms
Thread calibration: mean lat.: 1663.260ms, rate sampling interval: 6381ms
Thread calibration: mean lat.: 1610.518ms, rate sampling interval: 5992ms
Thread calibration: mean lat.: 1506.068ms, rate sampling interval: 5328ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 10.48s 4.99s 23.63s 65.29%
Req/Sec 3.49k 113.93 3.69k 55.88%
833028 requests in 1.00m, 277.26MB read
Requests/sec: 13884.09
Transfer/sec: 4.62MB
Note: It shouldn't be surprising that performance is less on AWS's memory-intensive instance class, because Ratpack is CPU-bound. It's worth keeping in mind that this is the case and that you need to do some work to find out the proper instance class for your application's use case.
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| c4.large | t2.medium | 15,373 |
# ./wrk -t4 -c72 -d60s -R20000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 1257.768ms, rate sampling interval: 5152ms
Thread calibration: mean lat.: 1201.606ms, rate sampling interval: 5210ms
Thread calibration: mean lat.: 1223.431ms, rate sampling interval: 5107ms
Thread calibration: mean lat.: 1079.383ms, rate sampling interval: 4616ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 7.88s 4.21s 19.32s 64.80%
Req/Sec 3.86k 79.60 4.04k 62.16%
922323 requests in 1.00m, 306.98MB read
Requests/sec: 15372.73
Transfer/sec: 5.12MB
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| c4.large | t2.large | 16,085 |
# ./wrk -t4 -c72 -d60s -R20000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 72 connections
Thread calibration: mean lat.: 938.818ms, rate sampling interval: 4073ms
Thread calibration: mean lat.: 943.311ms, rate sampling interval: 4079ms
Thread calibration: mean lat.: 1007.105ms, rate sampling interval: 4089ms
Thread calibration: mean lat.: 975.187ms, rate sampling interval: 4182ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 6.57s 4.36s 14.74s 60.80%
Req/Sec 4.03k 36.11 4.10k 70.21%
965082 requests in 1.00m, 321.21MB read
Requests/sec: 16085.08
Transfer/sec: 5.35MB
| EC2 Instance Type | RDS Instance Type | Max Req/Sec Recorded |
|---|---|---|
| c4.xlarge | t2.large | 23,149 |
# ./wrk -t4 -c100 -d60s -R25000 http://localhost:5050
Running 1m test @ http://localhost:5050
4 threads and 100 connections
Thread calibration: mean lat.: 346.475ms, rate sampling interval: 1235ms
Thread calibration: mean lat.: 347.531ms, rate sampling interval: 1232ms
Thread calibration: mean lat.: 347.625ms, rate sampling interval: 1254ms
Thread calibration: mean lat.: 347.428ms, rate sampling interval: 1245ms
Thread Stats Avg Stdev Max +/- Stdev
Latency 2.54s 1.13s 5.64s 59.59%
Req/Sec 5.80k 71.22 5.89k 83.02%
1388845 requests in 1.00m, 462.25MB read
Requests/sec: 23148.93
Transfer/sec: 7.70MB
Note: The first thing I'd like to note is that this combo went from a cold JVM to 20,000 rps without breaking a sweat. Normally, I run the test over 5 mins to ensure the JVM is warm (usually doesn't improve dramatically after the first round of calls), but this combo was seeing performance right out of the gates. The second thing I'd like to point out is that I increased the number of persistent connections to see if I could squeeze a few more RPS out of the run, so here you'll see 4 x 100 instead of my normal 4 x 72.