Hey everyone! I've been diving deep into the different implementations of the Gateway API lately and noticed a surprising amount of variation between them. After spending some time researching and benchmarking a few, I realized that the existing conformance tests don't fully capture how these implementations perform in real-world scenarios. They're great for simple checks, but when it comes to handling loads, upgrades, and other factors, there's much more to consider. Interestingly, only two out of thirty listed implementations have passed all conformance tests. I created a benchmark to evaluate various aspects like scalability and reliability. I'd love to hear your experiences and thoughts on these implementations! You can check out my [report here](https://github.com/howardjohn/gateway-api-bench) along with the testing steps. What are your suggestions for additional tests I should run?
3 Answers
Wow, thanks for the insights! I'm migrating to Envoy Gateway but wasn't aware of these benchmarks. After reading, I might lean more towards Istio, especially with all the upsides you mentioned. How do Envoy Gateway’s shortcomings compare with Istio, even though both use the same API?
Appreciate all your work on this! It sounds like you tested Cilium with the default settings, but did you consider its non-embedded option? I wonder how that would stack up against the other configurations.
I think the way the Envoy Gateway is set up isn't fully highlighted. There are different deployment options for the controller and proxy resources, including isolating them in namespaced deployment mode. It allows cluster operators to manage proxy resource creation more securely. Of course, it’s still in alpha stage, but it's good to know these options exist.
Related Questions
Sports Team Randomizer
10 Uses For An Old Smartphone
Midjourney Launches An Exciting New Feature for Their Image AI
ShortlyAI Review
Is Copytrack A Scam?
Getting 100 on Pagespeed Insights for Mobile is Impossible