Performance isn’t a nice-to-have – it’s a deciding factor for conversion, scalability, and whether your team can sleep peacefully at night. With Shopware 6.7, we put the platform through its paces under realistic load scenarios. At the same time, our partner 8mylez ran production-level tests with millions of data entries and consistently high traffic. The result: 6.7 scales – not just in theory, but under extreme data loads in real-world conditions.
In this article, you’ll discover how the tests were conducted, what performance gains 6.7 delivers, and what they mean for enterprise-scale operations.
What the benchmarks show (Shopware 6.6 vs. 6.7)
In our official Performance Whitepaper, we tested four scenarios – from organic traffic without caching to enterprise-level loads using Varnish and continuous API imports. You’ll find the detailed comparison tables for each scenario in the PDF (e.g., Scenario 2 on p.15, Scenario 3 on p.17).
Highlights from the results
Flash sale with Varnish (Scenario 2): Orders per second increased from 1.9 to 3.96 (+108%), while p95 latency dropped by up to 60%.
Enterprise load with API importers (Scenario 3): p95 latency fell from about 2 seconds to 321 ms; orders per second rose by 65% (2.29 → 3.79).
Frontend optimization (Sitespeed audit): Thanks to Vite and refactoring, JS/CSS payloads were reduced by ~25%, noticeably improving the performance score (76 → 78).
Why it matters
The benchmarks demonstrate that Shopware 6.7 performs far more consistently under high read/write concurrency – especially where Varnish and Valkey work together to reduce backend bottlenecks and ensure a stable, scalable system foundation.
Real-world proof: the 8mylez stress test with millions of data records
Our partner 8mylez, in collaboration with maxcluster, put Shopware 6 through a multi-stage load test – designed for millions of data records and continuous high load. According to their overview and results, the setup scaled reliably even under massive data volumes (see details and diagrams on pp. 3 and 8–10).
Setup and key figures
Data volume: up to 11 million products, 230 million customers, and 160 million orders (further increased manually – the system remained stable).
Load profile: approx. 600,000 requests/hour and 5,000 orders/hour, focusing on backend and server processes (frontend assets were intentionally excluded).
Response times: stable over more than an hour, averaging ~0.8 s, with p95 around 3.6 s, depending on cluster resources and configuration.
Tooling: A custom Locust-based load testing tool, built on Tideways open source, tailored to typical user profiles (see description on p. 7).
Best practices from the cluster setup (excerpt)
Load balancer and Varnish in front of multiple app servers
Redis (sessions, cache, cart, locking), RabbitMQ (queues), OpenSearch (search), and master/slave database setup
Shared filesystem for media and assets
Shopware best practices including xkey workarounds
Targeted PHP and database tuning (OPcache, InnoDB buffer pool, etc.) – see pp. 5–6 for details.
Key takeaways from practice
8mylez optimized caching for logged-in users and shopping carts, for example by implementing Ajax-based loading of personalized content and a more granular cache invalidation logic.
The result: higher cache hit rates and reduced server load – even with active, logged-in users (see p. 6).
Transparency about limits
For very large data volumes, areas such as Elasticsearch indexing, admin performance, backup/restore, migrations, and statistics can become challenging. Here, custom strategies – such as data archiving or external analytics solutions – are recommended to ensure long-term stability and maintain operational efficiency (see p. 9).
Theory × Practice: What this means for you
Shopware 6.7 delivers measurable improvements in throughput and latency across different load profiles (see scenario comparison tables).
8mylez proves that even under extreme data volumes, the system remains stably scalable – provided the architecture and caching layers are intelligently configured (see test phases and results).
In short:
Benchmarks provide planning confidence – for upgrade decisions, capacity management, and scaling strategies.
Real-world stress tests deliver hands-on insights into handling millions of data records in daily operations – including potential pitfalls and proven workarounds.
7 practical best practices for handling high load (ready to apply today)
1. Use HTTP caching consistently
Enable Varnish or Valkey and model your storefront caching strategies carefully. As shown in Benchmark Scenario 2, the impact can be significant – orders per second increased by 108%.
2. Cache even for logged-in users
Separate static from personalized content and load user-specific elements via Ajax. Manage invalidation rules precisely to prevent unnecessary cache bypasses – an approach proven effective in the 8mylez test setup (p. 6).
3. Scale horizontally – but strategically
Connect app servers horizontally in a clean setup, manage queues via RabbitMQ, and handle sessions and caching with Redis or Valkey. Assign search load specifically to OpenSearch/Elasticsearch to maintain stability under pressure.
4. Optimize your database
Review InnoDB buffer pool, I/O settings, threads, and binlog parameters (see the Performance Whitepaper appendix on MariaDB and DB tuning). Database tuning also played a key role in the 8mylez results.
5. Decouple indexing and reporting
For large catalogs, plan indexing strategies and consider external analytics tools. Avoid running statistical queries across the entire database – see page 9 for recommendations.
6. Plan migrations and backups early
As your data volume grows, migrations, updates, and backup/restore processes become more complex. Define archiving strategies and time windows in advance to minimize risk (see page 9).
7. Automate load testing early
Integrate load testing into your CI/CD pipeline to identify bottlenecks introduced by customizations before deployment – ensuring smooth scaling and stable performance (see page 9).
Who is this relevant for
CTOs and solution architects – responsible for making reliable upgrade and scaling decisions.
Dev and Ops teams – managing caching strategies, queues, search, and database tuning in daily operations.
Growing merchants (DACH region) – planning replatforming projects or preparing for rapid growth.
What’s next: webinar and resources
Together with 8mylez, we’re hosting a technical deep-dive webinar where we’ll answer your questions and show exactly how to apply the best practices discussed above in your own setup. The exact date of the webinar will be announced shortly. You will find more information here soon.
Shopware 6.7 Performance Whitepaper – all benchmarks, scenarios, and setup details.
8mylez Case Study (German) – stress tests with millions of data records, cluster blueprints, and performance optimizations.
