QWQyyy commented on issue #5410: URL: https://github.com/apache/openwhisk/issues/5410#issuecomment-1543308114
> The performance is highly related to the duration of your action. Let's say, your action takes 1 second to finish and you have 200 containers. Then your TPS will be just 200 TPS. If your action completes in 100ms, your TPS will be 2000 TPS. > > And you said, your machines have 32Gb of memory but you configured around 65GB of memory for runtime containers. It will exceed the limit of your machine and might cause OOM. Since core components like invokers should be running as well, you need to configure less than 32GB for runtime containers. And take a look at the duration of your activations. > > You can refer to [this guide](https://github.com/apache/openwhisk/blob/master/docs/intra-concurrency.md) for intra-concurrency. There are still some known issues with it though. Sorry, I have 3 nodes, so actually I have 32*3GB=96GB memory available, in K8s 2 work nodes have 64GB, and we also config it pool as 48GB, But it look like not good, Although the execution time of Action is very short, only 100ms,etc:  But the result is very confusing, My K6 test very unstable and poor performance, this is 300uvs and get 200TPS:  We also find some internal error:  But, I appreciate your comments, I will study the guide in the link carefully, I think I should turn to nodejs functions to support concurrent access, thanks! In the future, I will pay more attention to how to improve the TPS of OpenWhisk and solve some existing problems, which will be very helpful to academia and industry. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
