undertow

[Gustavo Galvan]

Hi people,

Somewhere in this group or github I read about using Undertow. Well... I was very curious about embedded undertow, then I tried with another micro-framework (which allows use jetty or undertow). The difference is amazing .... almost obscene.

I leave here the benchmark for a simple "Hello world" (ab -c10 -n10000 -k http://127.0.0.1:8081/)

Jetty:

Server Software:        Jetty(9.2.11.v20150529)

Server Hostname:        127.0.0.1

Server Port:            8081

Document Path:          /

Document Length:        12 bytes

Concurrency Level:      10

Time taken for tests:   0.709 seconds

Complete requests:      10000

Failed requests:        0

Keep-Alive requests:    0

Total transferred:      1410000 bytes

HTML transferred:       120000 bytes

Requests per second:    14100.59 [#/sec] (mean)

Time per request:       0.709 [ms] (mean)

Time per request:       0.071 [ms] (mean, across all concurrent requests)

Transfer rate:          1941.59 [Kbytes/sec] received

Connection Times (ms)

              min  mean[+/-sd] median   max

Connect:        0    0   0.1      0       2

Processing:     0    1   0.6      0       8

Waiting:        0    1   0.5      0       8

Total:          0    1   0.6      1       8

WARNING: The median and mean for the processing time are not within a normal deviation

        These results are probably not that reliable.

WARNING: The median and mean for the waiting time are not within a normal deviation

        These results are probably not that reliable.

Percentage of the requests served within a certain time (ms)

  50%      1

  66%      1

  75%      1

  80%      1

  90%      1

  95%      1

  98%      2

  99%      3

 100%      8 (longest request)

Undertow:

Server Software:        

Server Hostname:        127.0.0.1

Server Port:            8081

Document Path:          /

Document Length:        12 bytes

Concurrency Level:      10

Time taken for tests:   0.232 seconds

Complete requests:      10000

Failed requests:        0

Keep-Alive requests:    10000

Total transferred:      1510000 bytes

HTML transferred:       120000 bytes

Requests per second:    43022.41 [#/sec] (mean)

Time per request:       0.232 [ms] (mean)

Time per request:       0.023 [ms] (mean, across all concurrent requests)

Transfer rate:          6344.12 [Kbytes/sec] received

Connection Times (ms)

              min  mean[+/-sd] median   max

Connect:        0    0   0.0      0       1

Processing:     0    0   0.5      0      15

Waiting:        0    0   0.5      0      15

Total:          0    0   0.5      0      15

Percentage of the requests served within a certain time (ms)

  50%      0

  66%      0

  75%      0

  80%      0

  90%      0

  95%      0

  98%      1

  99%      1

 100%     15 (longest request)

[Christian Rivasseau]

Gustavo thanks for sharing,

Could you share the code? Last time I tried to embed spark on undertow I couldn;t figure

it out for some reason.

--
You received this message because you are subscribed to the Google Groups "sparkjava" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sparkjava+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Christian Rivasseau
Co-founder and CTO @ Lefty
+33 6 67 35 26 74

[Gustavo Galvan]

Hi Christian,

Maybe I wasn't clear, I've used another similar framework (because I can't embedded in spark too). I've benchmarked spark and the another framework (both with Jetty) and speed is the same. It is not my intention to advertise other project here, but for your info. the framework used for benchmark is Pippo.

[Christian MICHON]

I'm not sure the ab results are meaningful.

- yes undertow is asynchronous

- but jetty puts more info in headers etc... than undertow.

Look at it this way:

- Server Software is empty with undertow, is "Jetty(9.2.11.v20150529)" with jetty

- "hello world" is 50% of string "Jetty(9.2.11.v20150529)"

- so patching jetty for not sending Server Software information would make it ~3 times faster too?

I agree with Christian R: it'd be good to have both code snippets to compare, whether you're using spark/pippo or not.

For what it's worth, jetty without spark is even faster...

[Gustavo Galvan]

Code is only one because pippo support Undertow out of the box. Just only add the jars in classpath (Jetty or Undertow, and its dependencies), and pippo discover the server to use.

package test;

import ro.pippo.core.Pippo;

public class test01 {

public static void main(String[] args) {

Pippo pippo = new Pippo();

pippo.getServer().getSettings().port(8081);

                pippo.getApplication().GET("/", (routeContext) -> routeContext.send("Hello World!"));

                pippo.start();

}

}

[Gustavo Galvan]

Based on Christian Michon comment, I made a new benchmark now with payloads (and the recent addition of Tomcat embedded).

Code:

package test;

import java.nio.charset.Charset;
import java.util.Random;

import ro.pippo.core.Pippo;

public class test01 {

  public static void main(String[] args) {

  // generate payload
  byte[] array = new byte[1024*10];
  new Random().nextBytes(array);
  String payload = new String(array, Charset.forName("UTF-8"));
  // pippo stuff

  Pippo pippo = new Pippo();
  pippo.getServer().getSettings().port(8081);

        pippo.getApplication().GET("/", (routeContext) -> routeContext.send("Hello World!\n"+payload));
        //pippo.getApplication().GET("/", (routeContext) -> routeContext.send("Hello World!"));
        pippo.start();
  }
}

Benchmark command: ab -c100 -n100000 -k http://127.0.0.1:8081/

Results (req/s):

Server    | "Hello World!" |  payload (+10kb) |  payload (+5kb) |  payload (+1kb)

---------------------------------------------------------------------------------

Undertow  |    70659.84    |     9192.98      |    27928.34     |   52048.23

Tomcat    |    20315.36    |    10649.05

Jetty     |    19295.64    |    10255.17

Enjoy!

On Thursday, June 11, 2015 at 4:31:06 PM UTC-3, Gustavo Galvan wrote:

[Christian MICHON]

Thanks it's starting to be more meaningful yet some numbers seem to be missing so no valid conclusion is possible for now.

Thanks also for the pippo snippet.

[Gustavo Galvan]

I made one more to compare server degradation with different payload (just request/s):

Undertow

--------

             --- concurrency ---

payload      x1     x10     x100

---------------------------------

  1kb     11700   41000    51500

  5kb      6200   22200    25500

 10kb      3200    7500    11300

 25kb      1680    4700     5400

 50kb       937    2850     3190

Jetty

-----

             --- concurrency ---

payload      x1     x10     x100

---------------------------------

  1kb      7500   16400    17600

  5kb      5100   10200    12400

 10kb      3500    8400     9600

 25kb      2000    5400     5900

 50kb      1050    3400     3600

So, my basic conclusion is that undertow is the choice if your average responses are below 10kb. If not, has non sense.

But this is not a negative comment because I have a project with an API service that response average is below 10kb ! I could double the performance (and reduce costs) with Undertow.

On Thursday, June 11, 2015 at 4:31:06 PM UTC-3, Gustavo Galvan wrote:

[Gustavo Galvan]

UPDATE:

I configured Undertow's buffer size to 32kb (default is 16kb) and performance is up again, even with large data !!

It's a beast !!

On Thursday, June 11, 2015 at 4:31:06 PM UTC-3, Gustavo Galvan wrote: