This report benchmarks the performance of the Cloudways new Lightning stack against the older Hybrid stack across a variety of real-world WordPress use cases, including WooCommerce and LearnDash.
Optimized for Speed & Response Time
The Lightning stack consistently delivers faster response times across most test cases. In uncached and write-heavy workloads, where requests bypass full-page caching and access MySQL and persistent object caches like Redis, response times improved by up to 53% compared to Hybrid.
Performance Improvement
Overall, Lightning completed more critical actions per run and processed more successful requests per second on average. In LearnDash course list views, throughput was 85% higher, while WooCommerce saw 58% more successful checkouts under load.
Performance Under Higher Concurrency
Under high concurrency, Lightning handled more simultaneous users with less degradation in performance. Throughput improved by up to 46% for e-commerce workloads and up to 73% for high-concurrency LMS scenarios.
Faster Backends – Infrastructure Efficiency
Lightning also delivered a faster WordPress Dashboard experience, with up to 31% better response times for logins and up to 20% faster publishing workflows across all tested configurations.
Test Methodology
To evaluate performance objectively, we ran a series of load tests across 12 server configurations, evenly split between the Hybrid and Lightning stacks. Each server configuration ran 15 tests with 3-5 different concurrency levels. The tests fell into four categories:
- Locust: simulates real users performing actions, such as adding a product to cart, or completing a course lesson
- Lighthouse: for front-end and Core Web Vitals metrics
- Curl: checks server responsiveness and connection setup speed (including DNS, latency, TLS/TCP handshakes)
- The wpbenchmark plugin: for CPU, memory, disk and other tests from within a WordPress environment
Each test scenario was designed to mimic core segments of typical user flows:
- WordPress: Cached, Uncached and Static asset download
- WooCommerce: Cached, Uncached, Add to Cart and Checkout flows
- LearnDash: Cached, Uncached and Course progression
Each configuration was tested at various concurrency levels: starting at 200 users for cache and read-heavy workloads, down to 40 for write-heavy tests. Tests were repeated across four regions to account for network variability and simulate real-world latency. Consistent and predictable dummy data was generated on each target site, including posts, pages, media, products, courses, lessons and more:
| WordPress: Static | 10 posts |
| WordPress: Cached | 10 posts |
| WordPress: Uncached | 10 posts |
| WordPress: Login | 10 posts, 80-320 users, based on concurrency |
| WordPress: Publish | 10 posts, 40-160 users, based on concurrency + 1 posts per iteration |
| WooCommerce: Cached | 20 posts, 20 media, 20 products, 100 orders, 100 users |
| WooCommerce: Uncached | 20 posts, 20 media, 20 products, 100 orders, 100 users |
| WooCommerce: Cart | 20 posts, 20 media, 20 products, 100 orders, 100 users |
| WooCommerce: Checkout | 20 posts, 20 media, 20 products, 100 orders, 100 users + 1 order per iteration |
| LearnDash: Cached | 20 posts, 10 pages, 10 media, 10 courses, 10 users |
| LearnDash: Uncached | 20 posts, 10 pages, 10 media, 10 courses, 10 users |
| LearnDash: Lesson | 20 posts, 10 pages, 10 media, 10 courses, 40-160 users based on concurrency |
Each test ran for 20 minutes, with its peak concurrency reached in the first 10 seconds. No further ramp-up or cooldown periods were defined. Any requests waiting for a response at the end of the 20 minutes, were abandoned and disregarded from the final results.
We recorded the total number of successful critical actions (CA) for each test, and the average response time across them. We recorded all other available metrics on a per-region and per-URL basis, however our comparison was primarily based on successful CA metrics.
Server Configurations
All performance testing was conducted across 12 applications on 12 individual Cloudways servers. Half of the servers were configured to use the Lightning stack, while the other half remained on the default Hybrid stack. The server specifications for each stack were as follows.
| Provider | RAM (GB) | CPU Cores | Disk (GB) | Bandwidth (TB) |
| DigitalOcean | 4 | 2 | 80 | 4 |
| DigitalOcean | 8 | 4 | 160 | 5 |
| DigitalOcean | 16 | 8 | 320 | 6 |
| Vultr | 4 | 2 | 128 | 3 |
| Vultr | 8 | 3 | 256 | 4 |
| Vultr | 16 | 4 | 384 | 5 |
All tests were performed using the PHP 8.3 runtime, Breeze and Object Cache Pro plugins, with Varnish caching turned on. Additionally, a small subset of tests were repeated with Varnish caching turned off.
Additional plugins and themes used:
- WooCommerce 9.9.3 for WooCommerce tests
- LearnDash 4.21.4 for LearnDash tests
- WPBenchmark 1.6.2 for WPBenchmark tests
- Bench2 (our internal testing plugin)
- Twenty Twenty Five 1.2 for WordPress and LearnDash tests
- Storefront 4.6.1 for WooCommerce tests
Benchmark Results
The three cached tests (WordPress, WooCommerce and LearnDash) have demonstrated overall similar results across all concurrency levels, with little variance between Lightning and Hybrid stacks.
However, a few of the tests stood out, where Lightning underperformed or outperformed the Hybrid counterparts by a significant amount (-9% to up +12%) on mid-to-high concurrency levels. Given the inconsistent nature of these, it is likely this variance is caused by cache stampede, noisy neighbors, resource sharing/throttling or other external factors.
The uncached tests (WordPress Uncached, Login, Publish; WooCommerce Uncached, Cart, Checkout; LearnDash Uncached, Lesson) have shown a significant throughput increase for the Lightning stack versus the Hybrid variants, ranging from 1% up to 85%.
The biggest gains were observed in the more complex workflows with write-heavy workloads, such as WooCommerce Checkout, and LearnDash Lesson progression, for example:
- Lightning 8G has processed 4809 critical actions, versus only 3035 on the Hybrid stack on DigitalOcean (a 58% improvement) for the WooCommerce Checkout test at 80 concurrent users.
- Lightning 4G has processed 3123 critical actions, versus only 2034 on the Hybrid stack on Vultr (a 53% improvement) for the LearnDash Lesson progression test, at 160 concurrent users.
- Lightning 4G has processed 25031 critical actions, versus only 13459 on the Hybrid stack on Vultr (an 85% improvement) for the LearnDash Uncached test, at 80 concurrent users.
The biggest improvements (20%+) for these write-heavy tests were observed on lower-tier servers (4G, 8G), with the gap much closer (up to 10%) on high-tier configurations (16G) at the various tested concurrency levels. This is primarily because smaller servers have fewer total resources and are more likely to hit their performance ceilings under load. In these constrained environments, the performance difference between the stacks becomes more pronounced: Hybrid servers tend to reach their limits sooner than Lightning servers, as they are generally slower to process each request. This inefficiency compounds under load, resulting in a wider performance gap on lower-tier plans.
| Test | Min | Max | Average | Median |
| WordPress Cached (Varnish on) | -3.79 | 30.04 | 1.72 | -0.02 |
| WordPress Cached (Varnish off) | -6.70 | 209.23 | 60.14 | 17.46 |
| WordPress Static | -8.21 | 40.31 | 6.29 | 2.45 |
| WordPress Uncached | -2.26 | 45.61 | 21.76 | 24.90 |
| WordPress Login | 11.13 | 47.62 | 27.96 | 25.00 |
| WordPress Publish | -3.33 | 44.61 | 22.98 | 27.00 |
| WooCommerce Cached | -0.28 | 1.28 | 0.04 | -0.07 |
| WooCommerce Uncached | -0.88 | 49.06 | 21.93 | 19.56 |
| WooCommerce Cart | -0.08 | 56.07 | 25.56 | 25.02 |
| WooCommerce Checkout | -2.32 | 58.45 | 23.25 | 24.25 |
| LearnDash Cached | -0.69 | 0.62 | -0.06 | -0.12 |
| LearnDash Uncached | -1.18 | 85.98 | 25.82 | 18.67 |
| LearnDash Lesson | 1.65 | 53.54 | 26.79 | 30.27 |
Our Curl and Lighthouse/CWV tests have shown no significant (and consistent) difference between Lightning and Hybrid stacks. Furthermore, in both variants and as expected, latency and speed were significantly better for workers located in regions closer to the target servers (N. Virginia and São Paulo versus London and Sydney).
The WPBenchmark plugin tests lacked consistency between runs, suggesting external factors and/or instability in the tests being run. Furthermore, the Object Cache tests failed randomly across multiple server configurations, causing frequent reruns. It seems that Vultr has slightly outperformed DigitalOcean as the backing provider in most metrics, likely due to the difference in resource allocation, however, given the unreliable nature of these numbers, we are unable to make any claims.
Impact of Varnish
We ran several test cases with Varnish caching disabled. The Hybrid stack highlighted a strong dependency on Varnish, while Lightning-based applications performed only slightly worse without Varnish at high concurrency. For example:
- Hybrid 4G has processed 184k critical actions in a WordPress cached test with Varnish turned off, compared to 654k with Varnish enabled, on DigitalOcean at 1000 concurrent users. That’s a 71% degradation in throughput.
- Lightning 4G has processed 572k critical actions in the same test with Varnish turned off, compared to 619k with Varnish enabled. That’s only a 7% degradation.
Below is a chart demonstrating the effect of Varnish on total requests on Hybrid and Lightning stacks at 1000 concurrent users.
Overall, applications with Varnish enabled tend to perform better at high concurrency. At low concurrency however (200 users), Hybrid stacks are slightly (but consistently) better off with Varnish disabled.
The reduced reliance on Varnish in the Lightning stack has clear benefits for real-world WordPress sites. While Varnish excels at serving cached pages, it can limit performance when handling dynamic content, personalization, or real-time updates, which bypass full-page caching.
Lightning handles uncached requests more efficiently, improving performance for logged-in users, WooCommerce checkouts, LMS progress tracking, and other personalized features. This allows site owners to build more dynamic, interactive experiences without complex caching workarounds or performance tradeoffs. It also lowers the risk of cache misses during traffic spikes, making dynamic applications more reliable at scale.
Other Observations
We’ve noticed a pattern that’s consistent across the majority of our tests: there seems to be a consistent performance drop on the 0th second of every minute. This includes both cached and uncached tests, as well as our static test for images, JavaScript and CSS assets.
Below is a graph demonstrating this on our WordPress Static test with 1000 concurrent users, on an 8GB Vultr Lightning stack:
The Hybrid stack shows a very similar pattern. Other tests, especially the write-heavy tests, are less evenly spread out, however the peaks/drops are still quite consistently at the 0th second mark.
We’ve additionally verified this behavior using a separate, unrelated to our benchmarking framework, jMeter load testing instance, and an independent Loader.io test, and have observed the same pattern. jMeter:
Loader.io (time is relative to test start, peak is at exactly :00):
This behavior can be caused by various things, including cache expiration and/or invalidation, a cron job running on the 0th minute and other things. While this problem does not impact the Lightning vs. Hybrid performance directly, solving it will further improve overall platform performance, especially on cached and static delivery.
Conclusion
The benchmarking results establish clear performance baselines and demonstrate Lightning’s superiority across all critical metrics.
The Lightning stack consistently outperforms the Hybrid stack in real-world WordPress scenarios, especially under higher concurrency and in uncached or dynamic workloads, for example:
- Lightning achieved 21% better throughput on average in uncached workloads
- Lightning achieved 23-27% better performance on average in write-heavy workloads
While both stacks perform similarly in cached tests, Lightning shows clear advantages in responsiveness, throughput, and reliability without depending heavily on Varnish. These improvements are most noticeable on lower-tier servers and in write-heavy use cases like WooCommerce checkout and LearnDash course progression.
For most workloads, Lightning offers better overall performance and a more robust hosting environment.
Benchmarks and analysis performed by: Gennady Kovshenin, Karl Kubelet. Raw data available here. Koddrio, Inc. is a small distributed team of consultants with over three decades of combined WordPress experience. We help hosting providers and agencies build faster, more secure, and more scalable WordPress platforms. If you have any inquiries about these benchmarks, please feel free to reach out to us at hi@koddr.io.