📜 ⬆️ ⬇️

Load testing from the cloud

Any more or less serious web project, be it a portal, a service, an online store or something else, needs load testing, since it is this that allows you to identify potential “weak points” and errors in the code, and also makes it possible to evaluate whether the decision together with the platform to withstand the planned size of the audience of the project. You are planning to launch and it should be successful, right?



You can organize load testing yourself, in local mode, using your own server equipment and customer or developer infrastructure, or you can go the other way and test the project using cloud solutions. The cost of load testing from the cloud greatly benefits classical methods.

One of the most labor-intensive moments in the organization of the load testing process is the development of a load testing profile, which should include the necessary number of site usage algorithms, several types of users, etc. That is, to simulate the variability of site use, as in life. This is the task of analyzing how the site will be used, is in itself complex, but the question is rather creative and we are not talking about it now. Suppose we coped with it and made a list of scenarios for using the site. A large or small list, but all these scripts need to be turned into test scripts that need not only to be written, but also kept up to date. After all, testing is carried out more than once, we start the project seriously, therefore, every time after a change, revision or improvement of the program code of the project, we turn to testing. All this requires a parallel modification of test scripts.
')
When a testing specialist writes such a script, HTTP traffic capture technologies are used more often. For the script to work correctly, the specialist must bind it to the server parameters, which are taken from the script interaction session with the server. This session is usually called a virtual user. In order for each new virtual user to use the correct session parameters when entering the server, you need to develop scripts in such a way that they can take different parameters from the server and reuse them throughout this session.

But when the application has a lot of server-side parameters, this task becomes very laborious, because the tester has to maintain changes in the script at the code level or rewrite the script at all, and then modify the parameterization.

Hewlett Packard Enterprise offers a StormRunner Load SaaS solution that can emulate users at the browser level using TruClient technology. In this case, the script is not a stream of http-traffic, but a sequence of clicking on the buttons and menu in the browser. Monitoring server parameters in such a case becomes unnecessary, the browser follows them.
When you run the test, obviously, a browser is launched for each virtual user, that is, serious resources are consumed, but fortunately, StormRunner Load allows the customer not to worry about how much hardware capacity he needs to do load testing, the product takes the task of managing the test resources to myself. Thus, the headache with the search for resources for the load test is removed. At the same time, the funds allocated for the purchase of equipment are saved, if, of course, the company can afford it.

Any user, even very far from web programming, can independently record the TruClient script, since all you need to do is turn on the record button and independently play the specified sequence of steps in the browser that this script will perform. Of course, TruClient technology supports working with script parameterization to use various test data. Such a script is much easier to parameterize and modify in the application interface itself, and it is no longer necessary to rewrite it from scratch, which we have already mentioned above.

The HPE StormRunner Load solution during internal testing by HPE was tested on one and a half million simultaneous connections to the server under test. This is sufficient for most load testing tasks, even in large projects. But even if you need to test a web resource on a hundred or a thousand simultaneous connections, the product will be convenient to use at a reasonable cost.

In what practical problems can HPE StormRunner Load help? This may be a site check for reliability before starting a marketing campaign, when there is a rush of interest in the resource. If this campaign is successful and causes massive republishing, resource owners may experience habraeffect. In this case, users simply “tumble down” the site. It is also possible that the business that owns the site on certain days experiences peak loads. This could be a bank offering very interesting and favorable deposit terms, a store announcing another very profitable share, a news resource, publishing some sensational information, etc. The decision on load testing in all these cases can help to simulate the predicted load for the web resource, reproduce it in advance and check how the resource behaves.

If you wish, you can not only check the site’s stability with the predicted number of sessions, but also calculate the maximum level of load that a web resource is able to withstand. To infer how the test passed, user pass analysis is used - a sequence of user actions in the interface. In essence, a user transaction passing time is a metric of a site’s successful or unsuccessful load testing. You determine in advance how quickly a user transaction should complete, and the StormRunner Load task to collect statistics on these user transactions. Finally, the solution allows you to emulate the type and quality of the network from which test virtual users “log in”. This is relevant for those cases where the company would like to adapt its project for owners of different types of devices, including mobile, with different speed and time delay of the network channel.

Test using HPE StormRunner Load can be quite complex web applications, such as, for example, Internet banking systems. Their work is connected with integrations affecting a number of other banking systems and applications (ABS, credit conveyor, cards, CRM, and others). In this case, the customer can independently determine what time should be allotted for each user transaction and lay it as an SLA. This will allow us to determine at what stage the delays in processing the user transaction arose and additionally, using the collected hardware and hardware performance metrics, to analyze the root cause of the delay.

The HPE StormRunner Load solution is housed in the data center of Hewlett Packard Enterprise and architecturally consists of two parts. The first is on the HPE side and is responsible for user interaction and storage of all its information, test environments and scripts. The second part of the infrastructure is deployed in data centers, where there will be machines for load testing. This cloud Amazon, Microsoft Azure, the customer can choose the one that suits him. The customer can also use their own load stands, located in their data center or server.

Today, current HPE customers use classic products such as HPE LoadRunner and HPE Performance Center, which are deployed on their territory, in the data centers of companies and successfully solve the problems of load testing of any complexity for solving load testing problems. The new HPE StormRunner Load solution solves the problem of external testing under high loads and removes the question of finding equipment for the stand.

Source: https://habr.com/ru/post/276663/


All Articles