The world of orchestration and automation often promises "high performance," but performance is relative. At CodeZero, we believe that efficiency is not just a technical detail, it is the foundation of a scalable business. To see how we stack up against the best in the industry, we put CodeZero head-to-head against Kestra.
To ensure the results were undeniable, we used an Apple M3 MacBook Air and built the exact same workflow on both platforms. The result is a wake-up call for anyone who thinks enterprise orchestration has to be resource-heavy.
The setup: identical logic, no excuses
A benchmark only works if the playing field is level. We mirrored the same logic on both CodeZero and Kestra: a simple HTTP webhook that receives a request and returns a 200 OK response. By stripping away external API calls and database delays, we focused solely on the execution engine's raw speed.
For this test, we capped resources at 1.0 vCPU and 4GB of RAM to simulate a standard production instance.
Throughput: 69x more work on the same hardware
Throughput is the ultimate measure of how much your system can handle before it collapses. If your engine spends more energy managing itself than executing your tasks, you hit a performance wall.
The data is staggering. At a peak load of 200 virtual users, CodeZero maintains a massive 2,147 req/s. In the exact same environment, Kestra drops to just 31.05 req/s. This means CodeZero is doing the work of 69 Kestra instances on a single core. When you scale your business, this difference is the gap between a $100 monthly server bill and a $7,000 one.
Latency: why Kestra keeps your users waiting
Latency is where the end-user feels the weight of the architecture. High latency doesn't just feel slow, it breaks real-time webhooks and causes timeouts in critical business processes.
While CodeZero responds in a rock-solid 0.1s, Kestra’s response time spikes to 1.7s under load. Kestra is effectively 17x slower than CodeZero. In a world where every millisecond costs money, waiting nearly two seconds for a simple webhook response is a total system failure. CodeZero ensures your automations feel like raw code, providing instant feedback regardless of the load.
Why the gap is so large
Since the workflows were identical, the 69x difference in throughput isn't about the logic. It is about the language.
Legacy tools like Kestra are built on the JVM (Java Virtual Machine). While powerful, these platforms carry massive internal bloat, require "warm-up" periods, and consume significant memory just to stay active. CodeZero, however, is built from the ground up in Rust.
By using Rust, we’ve eliminated the typical "no-code tax." Our engine offers:
- Zero-cost abstractions: We get the safety of no-code with the raw speed of a low-level language.
- Minimal memory footprint: CodeZero is up to 160x smaller than traditional engines, leaving the RAM for your data, not the engine's overhead.
- Fearless concurrency: Rust allows us to handle thousands of requests simultaneously without the context-switching bottlenecks that cause legacy tools to choke.
