Got -ve min time using Netty Benchmark

[243826]

2015-09-14 17:57:37,675 [nioEventLoopGroup-1-0] INFO  netty.EchoTcpServer channelRead - Received the first message.
2015-09-14 18:02:00,407 [nioEventLoopGroup-1-1] INFO  netty.EchoTcpServer channelRead - Finished receiving messages! Overall test time: 262720 millis
results=Iterations: 1000000 | Avg Time: 65.193 micros | Min Time: -288764.0 nanos | Max Time: 4.22 millis | 75% Time: 71.975 micros | 90% Time: 79.043 micros | 99% Time: 109.072 micros | 99.9% Time: 154.557 micros | 99.99% Time: 647.032 micros | 99.999% Time: 2.314 millis

How is this possible?

[vrozov]

One possibility is that we timestamp message using System.nanoTime() and use it to compute time for the server to receive and process it. As System.nanoTime() is not guaranteed to be the same across different VMs even on the same machine, it is possible to get negative time difference.

[243826]

It's the same vm unless you mean jvm.

On Mon, Sep 14, 2015 at 6:43 PM, Vlad Rozov notifications@github.com
wrote:

One possibility is that we timestamp message using System.nanoTime() and
use it to compute time for the server to receive and process it. As
System.nanoTime() is not guaranteed to be the same across different VMs
even on the same machine, it is possible to get negative time difference.

—
Reply to this email directly or view it on GitHub
#24 (comment).

[vrozov]

Yes, I mean jvm.

[243826]

It's unbelievable!😰

On Mon, Sep 14, 2015 at 8:10 PM, Vlad Rozov notifications@github.com
wrote:

Yes, I mean jvm.

—
Reply to this email directly or view it on GitHub
#24 (comment).