Pushing performance limits – Shopware stress test with millions of orders

A large online shop in the healthcare sector planned the migration to Shopware 6—with millions of products, customers, and orders. As part of a pitch, load tests were carried out by the agency 8mylez, while maxcluster provided the highly available hosting infrastructure. The goal of the project was to find out how performant and scalable Shopware 6 can be operated under realistic enterprise conditions with millions of data records in a professional cluster setup.

Key results:

Shopware 6 is capable of running even very large shops with millions of items and orders with strong performance. The tests proved: With the right setup, even extreme load scenarios can be realistically replicated—reliably and with high performance.

• 99.99% availability with flexible scaling, even under peak load. Despite 230 million customer records and 160 million orders, the Shopware instance remained responsive and stable at all times.
• Up to 5,000 orders per hour and millions of items processed with strong performance.
• Up to 600,000 requests per hour were processed by the system.

Our Platinum partner 8mylez approached us with a special challenge: A prospective customer in the healthcare sector was planning to migrate a large online shop to Shopware 6 

The challenge was not only an extremely high data volume —with millions of product, customer, and order records that must be processed reliably and with high performance—but also whether the entire technical infrastructure, including hosting, can reliably support these requirements 

The key question was: Is Shopware 6, together with a modern, scalable hosting architecture, capable of meeting these requirements? And where are the technical limits of the system and the infrastructure when dealing with data volumes in the millions and heavy traffic?

To find out, 8mylez initiated a hands-on load test—focusing on high data load and intensive request load, for example to simulate customer orders with a later total volume in the millions while the system remains stable 

As the preferred hosting partner for complex e-commerce environments, we supported 8mylez in this hands-on load test with a specially configured test cluster, our Shopware expertise and many years of experience hosting demanding platforms.

Cluster setup for Shopware

For the planned test scenario, we provided 8mylez with a modular cluster setup specifically tailored to the requirements of a data-intensive Shopware system 

No chance of data loss and downtime

To ensure shops remain accessible even in critical moments—such as sudden load spikes caused by marketing campaigns, TV appearances, or seasonal peaks—maxcluster uses a fail-safe high-availability architecture in customer projects with an active and a passive server, redundant NVMe SSDs, and automatic fault detection—fully preconfigured and with no additional effort required from the agency or operator.

Identify performance issues before they occur 

In real customer projects, shop operators and agencies also benefit from maxcluster’s proactive 24/7 monitoring: more than 80 system and application parameters are continuously monitored—supplemented by automated shop and web checks. This allows potential performance issues to be detected and resolved early, before they negatively impact the shopping experience 

Hand over responsibility 

The best part: maxcluster takes full responsibility for monitoring, relieving agencies and merchants. In critical situations, Shopware and Magento experts are on hand to provide advice and deliver fast, targeted support.

A future-proof cluster architecture that grows with you and your business

Building on this high-availability and operations foundation, the setup architecture for the test case was designed so that key tasks such as load balancing, caching, database access, and queue handling could be clearly separated and implemented as dedicated components.
A central load balancer routed requests via Varnish to multiple application servers, each equipped with the full Shopware codebase and all relevant configurations. The environment was further complemented by powerful services:

• Redis for session handling, caching, shopping carts, and locking
• RabbitMQ for queue management
• OpenSearch for high-performance, scalable search
• A primary/replica database setup to decouple and optimize read and write access

Shared data—such as media or generated assets—continued to be provided via a shared filesystem. The configuration followed Shopware best practices, such as the targeted use of Redis in multiple areas and Varnish as a full-page cache—including the necessary workarounds for known limitations in the xkey implementation.

The initial setup was carried out on a single application server. The setup was then distributed to additional servers via a deployment mechanism and data synchronization.

Integration of third-party software 

In addition, an internal performance plugin from 8mylez (EmzPerformance) was used to further optimize the application layer in a targeted way. We supported the coordination of central cluster components—for example, the Redis setup for sessions—and ensured that the entire environment remains performant and stable even under high load. Targeted cache optimizations noticeably increased cache hit rates, leading to a significant improvement in speed and reliability—even under heavy load.

In addition to the caching adjustments, we provided targeted tuning at the PHP and database level:

• Review and optimization of the PHP version in use as well as relevant PHP settings
• Fine-tuning of the PHP OPcache for more efficient code execution
• Optimization of the InnoDB buffer pool and validation of the MySQL version for high-performance processing of large data volumes

The result: significantly improved interaction between the application and the database—a key prerequisite for efficient processes as user numbers grow and shop structures become more complex.

For the load tests, 8mylez used a custom-developed tool based on the Python framework Locust. As the hosting partner, we provided the necessary infrastructure and ensured a stable, high-performance environment for the intensive testing phases.

The test environment was dynamically scaled by maxcluster—resources such as CPU, RAM, and storage could be scaled live as needed—without any downtime or modifications. This made it possible to simulate load behavior realistically without being held back by static limits.

The tests were carried out in stages with constant traffic—up to 600,000 requests and 5,000 orders per hour were processed. The focus was on backend performance and the underlying server processes. Frontend resources such as images, JavaScript, or CSS were deliberately excluded, as they typically do not cause bottlenecks in such tests.

The load-testing tool used was developed based on an open-source project by Tideways and extended with specific Locust logic for Shopware. Since the standard version did not meet the requirements, 8mylez optimized the tool specifically for the Shopware context.

It enables fine-grained configuration of parameters such as user types, number of visitors, scenarios, and their weighting. In the tests conducted, 125 concurrent visitors were simulated over a period of one hour. The focus was on realistic customer traffic—API imports were not part of the test series, but could easily be added.

The load tests were carried out on a dedicated test cluster provided by us, in three clearly defined phases:

• Phase 1: First, a Shopware instance was set up without any data. This was followed by the import of 100,000 products as well as initial customer and order data.
• Phase 2: The number of products was increased to approx. 200,000 products  doubled, supplemented by 4,000,000 customer and order records each.
• Phase 3: The test data was further expanded to 600,000 products, 15,000,000 customers and  14,000,000 orders.

In all three phases, the system remained consistently stable and performant.

In addition, manual tests were conducted with a significantly expanded data set – with up to 11,000,000 products, 230,000,000 customers and 160,000,000 orders. Even under this massive data load, the Shopware instance remained stable.

The main goal was to ensure that the shop remains reliably accessible and performant despite extremely large volumes of data. Particular attention was paid to server response times, especially Time To First Byte (TTFB)* – both on average and at the 95th percentile.

Although 8mylez had not defined any fixed thresholds, measurements were systematically taken to determine how many requests and orders per second could be processed over a one-hour period without any noticeable drop in performance.

*The time until the server sends the first data packet.

Stable performance under high traffic and large data volumes

The load tests carried out by 8mylez on our test cluster clearly show: Performance remained consistently stable even under heavy load. During the one-hour tests, the number of requests and orders per second remained stable—just like the server response times.

The average TTFB was around 0.8 seconds, the 95th percentile at around 3.6 seconds —even under heavy data load and high user activity. Note: These values were measured at full utilization and on a highly data-driven setup—under normal operation they would be significantly lower.

Key findings from the tests

• The maximum number of simultaneous requests and orders per second depends directly on the resources available in the cluster.
• maxcluster’s hosting infrastructure proved to be powerful, stable, and flexibly scalable.
• Close coordination with the 8mylez team enabled fast and efficient implementation of the setup—including targeted technical advice on configuration and performance optimization.
• Shopware 6 provides a solid foundation for horizontal scaling: additional app servers can be integrated easily as traffic grows.

Interim conclusion

Large data volumes are not a challenge for Shopware 6 in an optimized environment. Even with over 200 million records, there was no performance degradation—neither when registering new customers nor when placing orders.

Where Shopware reaches its limits – and what can still be improved on the infrastructure side

As part of the load tests, it became clear: the key challenges lie less in Shopware itself and more often arise from technological limitations of adjacent systems—especially when very large volumes of data have to be processed 

Our partner 8mylez has identified the following problem areas for which individual technical solution approaches will need to be found in the future.

• Product indexing with Elasticsearch: As the amount of data grows, the indexing process takes increasingly more time and becomes a bottleneck with very large product catalogs.
• Performance in the admin area: Shopware’s standard admin interface reaches its limits when managing large volumes of data. While plugins exist to extend it, they are often not designed for use in complex enterprise setups.
• Migrations & updates: System and database migrations become more technically demanding with large volumes of data. Strategically planned processes are needed to implement updates in a performant, low-risk way.
• Statistics & reporting: Shopware’s standard reports access large data sets directly, which leads to noticeable delays with millions of records. Potential solutions could include integrating Elasticsearch or using specialized reporting tools.
• Caching & Varnish: Under heavy load, problems can occur if cached requests are not forwarded correctly to the app servers. Particular caution is required here, especially with Varnish configurations.

Fast setup and best practices

Although this project is a practical test case, our many years of experience with customer projects and our exchange with partners show that setting up a comparable setup with other hosting providers often takes significantly longer.

With maxcluster, server configurations are ready within minutes because the components are optimally coordinated and proven processes are used. This means projects are ready to go up to 80% faster than with other providers such as Hetzner or platform.sh. This saves valuable time and resources that can be invested in the actual development and optimization of the shop.

The load tests carried out by 8mylez on the test cluster we provided show impressively: Shopware 6 remains stable, powerful, and flexibly scalable even under heavy load. Even with massive traffic and large volumes of data, the platform stayed reliable and performant.

This was made possible by the perfect interaction with our specialized hosting infrastructure, which delivers modern technologies and best practices out of the box, creating the foundation for consistently stable performance.

Our key findings at a glance:

1. High scalability under load

Even at 5,000 orders per hour and sustained traffic with up to 125 concurrent visitors, there were no performance drops. The modular cluster setup proved to be flexibly expandable—a major advantage, especially during seasonal peak loads such as Black Friday.

2. Efficient handling of large volumes of data

In the tests, Shopware 6 processed several million records without any issues, including up to 600,000 products and 15 million orders. Through the targeted use of Elasticsearch and optimized caching strategies, consistently high performance could be ensured even for complex search queries and large catalogs.
Shopware 6 is therefore suitable not only for classic shop projects, but also for large, data-intensive e-commerce initiatives with high traffic. Combined with our scalable hosting architecture, the platform provides a robust foundation for future-proof, high-performance online retail.

3. Technical architecture as a success factor

The architecture of the cluster used—consisting of specialized components for load balancing, caching (Redis), queue handling (RabbitMQ), high-performance data access via OpenSearch, and a primary/replica database setup—was largely responsible for the results achieved.

The tests clearly demonstrate: With a technically mature infrastructure, Shopware 6 can be successfully scaled even for extensive, high-growth projects.

A particular advantage was the high flexibility in server configuration. Services such as Redis, Varnish, Elasticsearch, or OpenSearch are already included free of charge with maxcluster. This meant bottlenecks could be compensated quickly and a smooth test setup ensured.

We are pleased that we were able to support 8mylez as a hosting partner for this demanding test project—with a high-performance test cluster and targeted technical consulting.