Since it is part of a bigger implementation, we tested if it has a measurible effect on the usual path and also verified if it is helping where it should help.
In the case, where there is a long running read operation in our Redis server, we saw that other reads not waiting and working at full speed.
In the other cases, like there are only writes against the server, we can not find a performance problem. To be exact, the more cpu usage is not comparable to the latency we already have. The network calls(even in local network) and all other works are already much more bigger bottleneck then the additional CPU cost we introduced here.
Since it is part of a bigger implementation, we tested if it has a measurible effect on the usual path and also verified if it is helping where it should help.
In the case, where there is a long running read operation in our Redis server, we saw that other reads not waiting and working at full speed.
In the other cases, like there are only writes against the server, we can not find a performance problem. To be exact, the more cpu usage is not comparable to the latency we already have. The network calls(even in local network) and all other works are already much more bigger bottleneck then the additional CPU cost we introduced here.