In any software development project it is critical to carry out test plans that guarantee that the performance requirements of the application are met. Performance tests are, from the most obvious perspective, a set of tests that allow us to measure the speed of execution of a series of tasks in a system, under certain conditions.
To achieve a good level of performance, it is essential that the tests begin at the beginning of the software development. Moreover, if we want the results to be as reliable as possible, our test environment should be as close as possible to that of production, and never cross it with that of development or other tests.
Performance tests serve to:
- Demonstrate whether the system meets the established performance criteria;
- Compare systems to evaluate which one works faster;
- Validate and verify system quality attributes: scalability, reliability, use of resources;
- Measure which parts of the system or workload cause the assembly to perform poorly.
In this article, we will learn why testing environment is important; explore the different types of software testing staging and present recommendations to build performance testing environment.
Table of Contents
- Why is test environment in software testing important
- Staging or pre-production environment for load testing
- Determining the stand correspondence coefficient
- Production environment testing on simplified configurations
Why is test environment in software testing important
A test environment, also referred to as a sandbox environment, is one of many factors that can help you optimize your new software. This is a parallel environment to a production environment, where you can test new applications, modules, migrations and data import configurations and train users without compromising your organization’s actual data or disrupting operations. While this can increase the cost and schedule of implementation projects, in the long run, avoiding problems and unexpected situations can result in significant savings in money and time.
As such, the test environment for any implementation plan should be part of the project. This is important for large businesses and new clients, especially if we’re talking about IT installations that partners are not yet familiar with, or the implementation of a module that is not covered by basic finance such as distribution or project management. Once the process is complete and the management solution is resolved, it is important to maintain the test environment as it can be reused for any further testing or implementation.
Test environments can also be used for migration of solutions. In fact, this step is important in the development of the system and can have a major impact on execution if it is not executed or integrated properly. The test environment makes it possible to better identify any issues that may arise during a migration and therefore, create solutions for migration to a production environment.
In essence, a test environment protects data and tasks during migration or implementation, prevents cost and delays, and highlights potential issues and ensures solutions are ready. This is an essential component of any business management system, ensuring that it works before implementing and modifying the production environment.
Staging or pre-production environment for load testing
When we are talking about staging for load testing, many questions arise.
We are going to discuss the theory and our practices in organizing such a staging.
In an ideal load testing model 3 things should fit the production environment: the load profile, the test data and the staging. As a matter of fact, most big organizations are following the rule that the test staging for load testing should be 100% identical to the production environment. Banks are complying to this rule for the mission critical systems. But maintaining such a staging is costly. Therefore staging often has less resources than production. When the staging has got not less than 60% resources comparing to the production, then the load testing results are valid for the production as well. When the percentage is smaller, then the test results can’t be applied to the production environment. It’s not possible to perform load testing on one architecture and then apply the results to another architecture. For example, let’s say that your production system runs IBM POWER9, but the load tests were performed on Xeon: the test results don’t shed a light on the production system load capabilities.
Even when you haven’t got a possibility to create a staging with infrastructure and architecture as on production, there is still an opportunity to perform load testing and identify bugs and problems before the release. In this case we use the comparison (also release) load testing, where we compare the maximal performance for 2 releases. We suppose that if the performance falls 10% in the tests than there will be the same change on production.
We’ll need to analyze many details and look at every test to discover the reason for the lowered performance and bottlenecks. If the bottleneck on staging has got an analogue on production, most probably the error will also be on production. To name an example, we have found a memory leak of 1 GB per day using a load test. On staging the memory limit was 8 GB, so the test didn’t fail immediately. On production 32 GB were available, so the memory leak was not visible, but could lead to problems after some time.
Sometimes we apply not only the comparison testing, but also the test to find the stand correspondence coefficient: how much less resources has the staging comparing to the production.
Determining the stand correspondence coefficient
Load tests are a subcategory of performance tests focused on determining or validating the performance of the characteristics of the application under test when it is subject to workload simulating the use it will have in production.
During the project we determine the maximal system performance on a separate load profile for staging and production. We use the results to calculate the stand correspondence coefficient that can be applied to make prognoses about the production performance.
The base formula to calculate the stand correspondence coefficient is listed below:
where Lmax Production is the maximal production performance and Lmax Staging – maximal staging performance.
Each parameter stands for load level, matched to the 100% from the current production profile (taking into account operation restrictions).
While employing the comparison analysis we take into account the following details:
- which hardware is used on staging and on production;
- which performance indicators are the closest tot the coefficient from the base formula;
- if the bottlenecks on staging and production are the same;
- if there are any additional bugs that can influence the testing accuracy.
A more precise stand correspondence coefficient can be developed in the load testing approach after receiving the detailed information about the client system and testing environment.
Production environment testing on simplified configurations
Often it is not realistic to design a staging with all dependencies comparable to the production. Then we are employing module testing, applying the load testing component-wise. So we determine the most important system part and substitute the connected systems with the mock-ups.
We take away the systems that are not generating much of load and substitute their load with a similar load from another component. Such changes on one hand allow us to simplify the staging configurations and to speed up testing, on the other hand there are certain simplification limits and some problems in the integration may be unnoticed.
Applications are becoming more and more complex with shorter development cycles which require new and efficient methods of testing and development. The performance of an application in terms of overall user experience is now the key factor in the quality of an application.
Therefore, a performance testing strategy needs to be implemented at an early stage within the project life cycle. First step: qualification of performance. It defines the test effort for the entire project.
A traditional approach to environment for qa testing would require the project to wait for the application to be assembled before starting to validate performance. In a modern project lifecycle, the only way to include performance validation at an early stage is to test individual components after each build and perform end-to-end performance testing after application assembly.
Performance Lab has served over 500 companies across all domains, from finance and healthcare to retail and technology since its inception. It is one of the pioneering software testing services in the industry. By trusting our expertise, your business can benefit from core software testing services.