Guozhang Wang commented on KAFKA-2607:

Thanks for sharing Ismael. This is a great read.

I vaguely remember in J7 the perf difference of nanoTime is indeed 
non-negligible, but I cannot find the source of my memory now...

> Review `Time` interface and its usage
> -------------------------------------
>                 Key: KAFKA-2607
>                 URL: https://issues.apache.org/jira/browse/KAFKA-2607
>             Project: Kafka
>          Issue Type: Improvement
>    Affects Versions:, 1.0.0
>            Reporter: Ismael Juma
>            Priority: Major
>              Labels: newbie
> Two of `Time` interface's methods are `milliseconds` and `nanoseconds` which 
> are implemented in `SystemTime` as follows:
> {code}
>     @Override
>     public long milliseconds() {
>         return System.currentTimeMillis();
>     }
>     @Override
>     public long nanoseconds() {
>         return System.nanoTime();
>     }
> {code}
> The issue with this interface is that it makes it seem that the difference is 
> about the unit (`ms` versus `ns`) whereas it's much more than that:
> https://blogs.oracle.com/dholmes/entry/inside_the_hotspot_vm_clocks
> We should probably change the names of the methods and review our usage to 
> see if we're using the right one in the various places.

This message was sent by Atlassian JIRA

Reply via email to