We're currently migrating our `Webflux` project to Co-routines. Loving the increased readability of ...

# spring

[Gary]

We're currently migrating our Webflux project to Co-routines. Loving the increased readability of co-routines and the simplicity of async/await, great job @sdeleuzeand team. But under the hood we're wondering should we expect any performance differences? If so in what circumstances is this likely to be evident?

[Peter]

I converted a webflux service to coroutines and it is now running in prod for a week. At peak times it is serving up to 4 million requests / hour and 2500 requests / second. The service is deployed in a docker container with 1GB memory, and 1 vCPU of an AWS t3a instance type. Memory usage has not increased since switching to coroutines. CPU usage has slightly increased from previously 60% to 70%-80% within peak times. As the switch also included the update to Spring 2.2, I cannot say whether the increase of CPU usage is due to the switch to coroutines.

[sdeleuze]

Thank for sharing. The slight CPU usage increase is likely coming Reactive to Coroutines conversion. I will have a look later if we can do more optimizations.

[Gary]

Thanks for the update. We'll do some profiling before going live and update here with some benchmarks.

[Gary]

Screenshot_20191206-153008.png