TestnScale.com

Cycle and Think Times in Faban


Table of Contents

Introduction
Think Time
Cycle Time
Modeling Delays
Cycle Time Annotation
Deploy and Run
Learning More

Introduction

In this tutorial, you will learn what is meant by Cycle Time and Think Time and how you can add these to your workload. We will use the same sampleweb workload that was created in the Creating your first Workload in Faban Tutorial. Please follow the steps in the tutorial to install the samplewebapp. The workload can also be downloaded but the application needs to be deployed and Faban started.

Think Time

In any interactive application, when the user is presented a screen full of data, there is a time interval that passes while the user reviews the data and decides what the next operation should be. This time is called the 'Think Time' and most load testing tools provide a way to emulate it. This allows one to test the application with an emulated user population that a real application is likely to see.
Think Times are typically used in closed systems i.e. systems in which a known user population interacts with the application. Standard benchmarks such as TPC-C© use this model.

Cycle Time

Cycle Time in Faban refers to the inter-arrival time between subsequent requests that arrive at a server. For internet-based applications where the user population is unknown, it makes sense to model cycle times (rather than think times).  Note that there is a subtle difference between using think times and cycle times. In a load test that uses think times, if the server slows down and response times increase, the think times are still generated in the same manner giving the server time to recover. But in a test that uses cycle times, if the response time is large, the cycle time still remains the same - the emulated user simply sleeps for a shorter time between requests (and in the worst case where the response time is larger than the cycle time, will not sleep at all) causing the load on the server to increase. This will quickly result in excessive load on the server causing increasingly larger response times.  This is not necessarily a bad thing - it can pin-point scalability problems in the server more quickly and help you fix them.

Modeling Delays

Think times or cycle times are typically modeled using a negative exponential distribution with a mean time that is reasonable for the particular operation. The distribution is truncated at 5 to 10 times the mean. If you truncate the distribution too early, you may not get the specified mean. You may have to iterate a few times to figure out the right values for your mean and max. times. The figure below shows the measured cycle time distributions from an Olio (a web2.0 workload) run. This workload uses a mean cycle time of 5 seconds  and truncates the distribution at 25 seconds.



Cycle Time Annotation

Enough of the theory. Let's see how to specify a delay between subsequent requests made by an emulated user in Faban. This is done by editing the Driver source code. Go to the directory containing the sampleweb installation.

$ cd sampleweb
$ ls
README                       license.txt
build                        nbproject
build.properties.template    src
build.xml                    webapp
deploy
$

Edit the driver source file in src/com/testnscale/sampleweb/SampleWebDriver.java and find the FixedTime annotation starting on line 54:

@FixedTime (
    cycleType = CycleType.THINKTIME,
    cycleTime = 0,
    cycleDeviation = 5
)

This annotation specifies a fixed think time of 0 seconds - basically, the emulated user will simply cycle back to the next request without any delay at all. The cycleDeviation property specifies the percentage deviation allowed for this parameter - Faban will automatically check if the actual delay times are within the specified range and print the results in the summary report.

Change this to a negative exponential distribution with a mean cycle time of 5 seconds as follows:

@NegativeExponential (
    cycleType = CycleType.CYCLETIME,
    cycleMean = 5,
    cycleDeviation = 2
)

By default, Faban will truncate the distribution at 5 times the mean (25 seconds for this case). If you want a larger maximum value, add a cycleMax property. Make sure you delete the FixedTime annotation. Save the file.

Deploy and Run

Since we modified the driver, we must re-build and re-deploy the workload. Check that your build.properties file has the right entries and your Faban master is up.

$ ant deploy
Buildfile: build.xml

init:

compile:
Compiling 1 source file to /Users/testnscale/sampleweb/build/classes

bench.jar:
      [jar] Building jar: /Users/testnscale/sampleweb/build/lib/sampleweb.jar

deploy.jar:
      [jar] Building jar: /Users/testnscale/sampleweb/build/sampleweb.jar

deploy:

BUILD SUCCESSFUL
Total time: 14 seconds

To test the workload,  the samplewebapp must be deployed in a web server as described in the Creating your first workload tutorial.

Learning More


All Rights Reserved. See Terms and Conditions.