Is performance testing required for cloud-based applications?

Does the cloud “solve” the performance problem?

There is a widely held conception that as Cloud resources can auto-scale, this will solve all performance issues, so is performance testing applications in the cloud a waste of money and effort?

More and more companies are moving to the cloud from on-premises solutions for the obvious benefits such as scalability, flexibility, business continuity assurance and lower operating costs.

One of the key benefits of ‘cloud’ is scalability. Scalability is the ability to increase or decrease IT resources as needed to meet demand.

The term ’Scalability’ is often used next to or even interchangeably with performance, even though they are not as much related as many of us think.

Scaling resources won’t automatically make your application run fast, stable, and meet your business requirements (such as Service level agreements – SLAs, user experience, etc.).

In fact, auto-scaling itself needs to be tested (as part of load testing) as auto-scaling can be prone to performance issues, such as cold-start issues, when new nodes are added.

Load Testing:

The load testing is a key component of performance testing that helps assess an application’s behaviour under different load conditions by simulating multiple users, including normal load, peak load, soak load, and stress load scenarios.

By conducting load tests, the project team can gather valuable insights into response times, throughput rates, resource utilisation levels under different load conditions and bottleneck details. Also, it helps to identify the breaking point of an application.

Cloud Load Testing:

Cloud performance testing measures the performance and scalability of Cloud-based applications under simulated real-world user traffic under different load conditions such as standard, peak, stress, etc. While migrating from on-premises to the Cloud, you need to compare the performance benchmarks across the environments to ensure SLAs and user experience are met.

Load testing provides you with this information and ensures that Cloud resources can handle expected workloads and meet performance requirements.

If you have an application or service where the code is inefficient (bad code) and you push it into an auto-scaled ‘cloud’ infrastructure without performance testing and tuning it, if you’re lucky enough and have no major code, configuration, or design issues, it may appear to work just fine.

However, the extra infrastructure needed in the cloud to compensate for the inefficient ‘design’, ‘code’ and ‘configuration’ will start to hurt financially on the extra monthly ‘cloud’ infrastructure costs.

Throwing hardware at an application inefficiency can be a costly exercise. “Cloud” doesn’t magically make bad code & configuration better.

The cloud can make things easier in some respects, however, it can also make it harder and more complex to monitor and manage. When a cloud provider handles your infrastructure, you may not always have access to monitor what’s going on under the hood.

Your ‘cloud’ server could perhaps handle 100,000 requests per second, but if the end-user still has to wait 30 seconds to display an output, then there is an issue.

Conclusion:

Moving from on-premises to the cloud does NOT mean the application architecture complexity has vanished, you’re just passing it to someone else.  You are still running your application in a complex environment.

Undoubtedly, migration to the cloud offers high scalability; however, it doesn’t necessarily mean that cloud platforms can handle any workload without performance issues.

Don’t ever forget, Performance is more than capacity/scalability.

Without scalability, performance will be negatively impacted as demand is increased, though scalability in itself will not solve performance issues as there may be other reasons for the poor application performance (such as ‘poor database design/indexing’, ‘database locking’ ‘network latency and bandwidth issues’, ‘memory leaks, resource constraints’, ‘inefficient algorithms and code’, ‘lack of caching and overuse of resources’, ‘incorrect use of memory pool, ‘incorrect configuration’, ‘integration issues with external systems’, ‘concurrency and synchronisation issues’, etc.).

Auto-scaling can be a great way to handle unpredictable workloads. But auto-scaling won’t do anything to address the responsiveness or stability of your application. In fact, it’s adding more architecture complexity (invisible), potentially making it harder to keep the latency down.

So, to answer the question posed, does the cloud “solve” the performance problem?

Clearly, the answer is a big NO.