Testing for performance

[Quantumuniverses]

Why the performances (in sense of time) of my application are so different
between normal run and test?

I'd need to measure my app performances (as time) during testing, but it
seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

[Antoine Roux]

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

[Sebastian Gozin]

I would also add that you can use the -war option to run the application in a more production like mode.

Finally, I think I'd prefer to deploy a specially prepared war and then run tests from a separate suite than the ones in grails to ensure I can test on a production like server environment rather than my laptop.

On 29 May 2012, at 19:32, Antoine Roux wrote:

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

[Quantumuniverses]

They’re integration tests and have the classic Grails test structure:

 

class MyTests extends GroovyTestCase {

   

    static transactional = false

    def myController = new MyController()

    def mydataService

 

    protected void setUp() {

        super.setUp()

        myController.mydataService = mydataService

    }

   

    protected void tearDown() {

        super.tearDown()

    }

   

    void testGeneration() {

        // my code

    }

 

}

 

 

"Antoine Roux" <[hidden email]> ha scritto nel messaggio news:[hidden email]...

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

 

[Quantumuniverses]

What do you mean exactly with -war option? Can I execute my app by STS specifing it runs as a war or something like this?

 

I can produce the .war file and place it into my web app so it can run under Tomcat, anyway for me it’s enough it can run by STS and measure the performances (time).

 

 

"Sebastian Gozin" <[hidden email]> ha scritto nel messaggio news:[hidden email]...

I would also add that you can use the -war option to run the application in a more production like mode.

Finally, I think I'd prefer to deploy a specially prepared war and then run tests from a separate suite than the ones in grails to ensure I can test on a production like server environment rather than my laptop.

 

 

On 29 May 2012, at 19:32, Antoine Roux wrote:

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

 

[Antoine Roux]

Integration tests have a different setup than you production environment: each test runs inside a transaction, and the transaction is rolled back after each test. This could add latence, depending on how you measure performance.
Also, the application is running with all the infrastructure of running the application from the grails command. This could consume more memory.

On Wed, May 30, 2012 at 9:11 AM, QuantumUniverses <[hidden email]> wrote:

What do you mean exactly with -war option? Can I execute my app by STS specifing it runs as a war or something like this?

 

I can produce the .war file and place it into my web app so it can run under Tomcat, anyway for me it’s enough it can run by STS and measure the performances (time).

 

 

"Sebastian Gozin" <[hidden email]> ha scritto nel messaggio news:[hidden email]...

I would also add that you can use the -war option to run the application in a more production like mode.

Finally, I think I'd prefer to deploy a specially prepared war and then run tests from a separate suite than the ones in grails to ensure I can test on a production like server environment rather than my laptop.

 

 

On 29 May 2012, at 19:32, Antoine Roux wrote:

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email

 

[Quantumuniverses]

I used transactional=false so that no rollback should be done, and the situation is near the production one.

 

Also I can’t explain why there’s a so big difference between the performances of STS launch and JUnit test launch (2 minutes Vs 15 minutes).

 

You say the app is running with all the infrastructure.. yes, in fact I eventually expect better performances with integration tests than STS ones!

 

 

"Antoine Roux" <[hidden email]> ha scritto nel messaggio news:CAJZe56Z=[hidden email]...

Integration tests have a different setup than you production environment: each test runs inside a transaction, and the transaction is rolled back after each test. This could add latence, depending on how you measure performance.
Also, the application is running with all the infrastructure of running the application from the grails command. This could consume more memory.

On Wed, May 30, 2012 at 9:11 AM, QuantumUniverses <[hidden email]> wrote:

What do you mean exactly with -war option? Can I execute my app by STS specifing it runs as a war or something like this?

 

I can produce the .war file and place it into my web app so it can run under Tomcat, anyway for me it’s enough it can run by STS and measure the performances (time).

 

 

"Sebastian Gozin" <[hidden email]> ha scritto nel messaggio news:[hidden email]...

I would also add that you can use the -war option to run the application in a more production like mode.

Finally, I think I'd prefer to deploy a specially prepared war and then run tests from a separate suite than the ones in grails to ensure I can test on a production like server environment rather than my laptop.

 

 

On 29 May 2012, at 19:32, Antoine Roux wrote:

Which kind of tests are you thinking of? Integration? Functional?

Do you have an example of a use case that takes significantly more time in test mode?


Antoine

On Tue, May 29, 2012 at 5:09 PM, QuantumUniverses <[hidden email]> wrote:

Why the performances (in sense of time) of my application are so different between normal run and test?

I'd need to measure my app performances (as time) during testing, but it seems not to reliable..



---------------------------------------------------------------------
To unsubscribe from this list, please visit:

  http://xircles.codehaus.org/manage_email