Load Testing Calculators

One of the most common questions we get is "how many virtual users do I need to simulate?" The obvious answer is "how many people do you expect to have accessing the your system simultaneously?". If you have this information, then you're done!

But a tester or project manager frequently has less specific information available, such as stats from a web log analyzer. The calculators on this page will help determine how many virtual users are required to accurately simulate the expected load on a website.

How many virtual users do I need?

Peak Hourly Visits and Average Visit Length

If you know the peak hourly visit rate (the number of unique visitors to the website in its busiest hour) and the average visit length (the amount of time, on average, each use spends on the website), enter them below for an estimate of the required Virtual User count.

Peak visit rate (visits/hour)
Average visit length (minutes/visit)
Virtual Users required1000

Peak Hourly Pages, Testcase Size and Duration

If you know the peak hourly page rate (the number of pages visited on the website in its busiest hour) and the size and duration of the testcase (number of pages and duration in minutes), enter them below for an estimate of the required Virtual User count.

Peak page rate (pages/hour)
Testcase size (number of pages)
Testcase duration (minutes)
Virtual users required750

How many rows of data do I need?

It's common for datasets to consist of data is only useable once per test. For example, you might be logging in as a user, completing and submitting a one-time form, and logging out. In such a case, you'll need to calculate how many rows of data you'll need over the course of the test. Also, you'll need enough data so that when it's split up amongst the load engines, each load engine has enough data to avoid running out if the load engines are slightly unbalanced.

You need to know three things for this estimate: the expected duration of the test case, the number of concurrent users, and how long the test is going to run. This calculator assumes a steady ramp for the duration of the test: if you need to ramp more quickly and then hold at a level for a longer period of time, calculate the rows of data necessary for the ramp period first, then for the load period. For details on the calculations, refer to this blog post.

Note that this calculator does not take into account dataset rows with a lifetime shorter than "test case" - if you're using multiple data rows per test case iteration, simply multiply the below results by the number of rows in use per test case. Also, if you expect some errors, then pad the dataset to account for those test cases terminating early.

Peak concurrent users
Test duration (minutes)
Testcase duration (minutes, round down)
Rows of data required1250

How much is the bandwidth for my test going to cost?

If you are planning on testing an externally hosted server, this calculator can be used to help you plan bandwidth charges consumed by your virtual users. This can be useful when prioritizing content on the site. For more information, please read this blog post.

Load Configuration

Peak concurrent users
Test duration to reach peak users (minutes)
Test duration holding at peak users (minutes)

Testcase Configuration

Testcase duration (minutes)
Approximate number of testcase attempts over the course of the load test
Browser upload Browser download
Each testcase attempt (bytes)
Over entire load test (GB):

Data Transfer Charges

Content hosting provider OUT IN
Rate ($/GB)
Estimated charges to content provider
Load engine hosting provider OUT IN
Rate ($/GB):
Estimated charges to engine provider
Estimated bandwidth cost for load test?.??
Estimated bandwidth cost per visitor?.??

Notice: this example is provided for illustrative purposes only, and should be carefully compared with your environment. Each environment is different, and you should consider other factors that may be relevant in your environment, such as tiered pricing, data transfer between edge nodes & origin servers, data transfer between hosted servers, or pricing for services not based on bandwidth consumption. Additionally, this model assumes that test users will not deviate from their normal course of behavior, and does not consider error conditions or changes or changes in page performance.

Resources

Copyright © 2014 Web Performance, Inc.

A Durham web design company