Go back to all articles

Creation of the Load Testing Profile

Jun 27, 2020
5 min read

A load profile is a set of operations whose intensity is chosen to create a load comparable to the production traffic patterns.

Load testing preparation is one of the most important steps in load testing services.

An incorrect load profile generates testing results unrelated to the production.

We recommend to begin the load testing preparation by considering the testing goals, such as:

  • determining the maximal production load – you’ll need to prepare the load profile that maximally loads the system.
  • evaluating how certain processes influence the system – it’ll be needed to prepare a load profile for the period, when the processes are running on production; their influence is tracked not the whole system, but to the operations that communicate to the inspected processes.
  • determining the operations that will be most difficult to optimize later – first the operations-candidates need to be considered, then one needs to prepare one or more load profiles to audit them.

and so forth.

Collecting Statistics

Time range

When choosing the time range for the analyzed statistics, you’ll need to include several productive load cycles—they will allow you to determine the typical load. We recommend using the statistics for a period smaller than a month.

The time range should not be too big: if you take the statistics for half a year, the final load profile will display the average load for the whole period, which is not correct when the load level is constantly rising. A large amount of data will also make the analysis more difficult.

You’ll need to take into account that the load differs between the months: in December, the load is normally higher, but in January and May, there are many public holidays. Therefore, the statistics for these months may not be as informative. In this case, we recommend comparing it to the statistics for the next and previous months to evaluate the deviation. Alternatively, it’s possible to enlarge the analyzed period.

Aggregation

Normally, the operation intensity is aggregated every hour (in case the intensity is more than 100 operations per hour)

Statistics Analysis

Days to choose for the load profiles

Most systems have more and less loaded periods:

  • day and night
  • weekdays and weekend
  • days of the week (on Monday the load is often higher than on Friday)
  • the load can be performed on certain days of the month
  • or in a custom way, such as on the first Thursday of the month

It’s necessary to determine days with the typical load for the desired load profile.

It is the set of days when the load is similar and as high as possible. The amount of selected days should be enough to form a typical load.

For example
One day in a month with much different load looks untypical and can’t be taken as load profile base – even when the load during this day is the highest.

But when there are several such days and the system implies that in certain days the load is the highest – then the profile should be based on these days.

Internet rush hour and the corresponding time range

For the chosen days we divide operations to average out the operation intensity for each hour.

The load profile base is the Internet rush hour – the hour when the sum load for operation is maximal.

It’s necessary to evaluate the time range when the relations between operations are similar to rush hour. The load profile will cover this period. For the remainder of the time, we need to create a separate load profile.

If the load profile is aimed to analyze certain operations, it’s recommended to make sure that the chosen days and hours have the maximal intensity in respect of the operations.

And the other way round, not very important operations that are highly intensive should be discarded.

Choosing operations to include in the load profile

Based on the operation intensity for the chosen rush hour, we need to choose a set of operations to include in the load profile.

  • To begin with, it’s advised to set the load profile precision. Normally it is about 90% or more.
  • It means that one needs to sort the operations in the descending order of intensity and include those of them in the load profile that in sum are producing not less than 90% of load.
  • It’s important to understand that not the percentage coming from each operation is important, but rather the sum percentage from selected operations. There can be operations in the load profile that compound less than 1% from the total load).
  • The following operations can be included in the load profile even though they are not as precise as required.
  • Critical operations – operations that should be included in the load profile independent from their intensity, because they are crucial for business and influence the system work.

If it is known that the main load on the operations occurs on certain days/hours, we advise making sure that these hours are included in the analysis. Based on these hours, one may create the load profile.

  • Operations whose inclusion in the load profile doesn’t require any resources.

Determining the operation intensity in the load profile

  • The intensity of each operation should be enough to perform the operation on each test step many enough times to verify the corresponding requirements.

Determining the operation intensity in the load profile

As a result of the previous steps, we’ve compiled a list of operations and the time range for system work that the load profile will cover. We include the highest intensity for each operation for the time range covered by the load profile to the final load profile. After that, we evaluate how well the resulting profile covers the intensity of the rush hour. The allowed value is about 10%. If the allowed value is overtaken, the time range covered by the load profile must be corrected. For executing large-scale distributed load tests efficiently, consider leveraging a JMeter cloud load testing tool to optimize the intensity assessment and ensure seamless scaling.

It follows that the final profile emulates the highest possible load.

Batch and atomic operations

Some operations can be performed atomically or sequentially. For example, in the banking domain, after registering a client, an account is created. Batch operations can form a load different from performing the operations atomically. In such a case, we first determine the needed intensity for the batch operations from the composed profile and then give the rest of the intensity atomically.

Comparing the Load Profile to the Previous One

After creating the profile, it’s advised to compare it to the previous profile if one is present.

The profiles are compared in the following way:

  • Comparing the amount of operations
  • Comparing intensities for each operation in the load profile

(it’s performed after taking the old load profile to the level of the new load profile: the intensity of each operation from the old profile is multiplied to such a coefficient that the amount of the old profile operations equates to the number of new profile operations.

The frequency of load profile update

The necessity of updating the profile depends on the testing system’s load. It’s recommended that the load profile be updated once a quarter.

Want to Learn More About Our Performance Testing Services?
Find out what’s included and how to start working with us.
Table of contents

Related insights in blog articles

Explore what we’ve learned from these experiences
6 min read

Top 5 JMeter Alternatives

top jmeter alternatives preview
Dec 20, 2024

It’s hard to find someone in the performance testing community who hasn’t heard of Apache JMeter. We love it for being open-source, free, feature-rich, protocol-friendly, and easily extendable. While JMeter remains a favorite, there are other tools that offer unique strengths and advantages. This article presents a comprehensive list of the top 5 JMeter alternatives, […]

5 min read

How to Load Test API: A Full Guide

how to load test API- a full guide preview
Dec 18, 2024

In today’s digital ecosystem, APIs form the backbone of diverse software applications, facilitating communication and data exchange for an interconnected digital world. However, as demand for these services grows, ensuring their robustness and ability to handle varying levels of traffic becomes crucial. This is where PFLB, a next-generation, cloud-based load testing tool, comes in. In […]

7 min read

Top 8 Gatling Alternatives Overview

Top Gatling alternatives preview
Dec 16, 2024

Gatling Cloud, a cloud-based extension of the open-source performance testing tool, is a powerful load testing solution with its benefits. Those include excellent scalability, no-code test builder, moderate price for virtual user hours (VUh), and numerous useful integrations. However, with its steep learning curve due to reliance on Scala/Java and setup (and overall) complexity, it […]

13 min read

Top 10 BlazeMeter Alternatives

top blazemeter alternatives preview
Dec 13, 2024

Over a decade ago, BlazeMeter reshaped the landscape of load testing by moving it to the cloud. Serving as a cloud-based execution platform for the well-known JMeter, it freed engineers from the burden of managing infrastructure and allowed them to focus solely on testing. The result was a significant reduction in operational complexity and a […]

  • Be the first one to know

    We’ll send you a monthly e-mail with all the useful insights that we will have found and analyzed