Measuring Camel Performance

Question

I'm suspicious of my times to process messages due to it varying widely regardless of number of consumers and numbers of threads.

My code is:

@Component
@Scope("prototype")
public class ThreadStopWatch {
    private long startCpuTime = -1;
    private long stopCpuTime = -1;
    public Long start() {
        startCpuTime = System.nanoTime();
        stopCpuTime = startCpuTime;
    }
    public getElapsedTime() {
        stopCpuTime = System.nanoTime();
        return stopCpuTime - startCpuTime;
    }
}
...
private ThreadStopWatch threadStopWatch;

public configure(...) {
    for(Endpoint endpoint: myendpoints) {
        if(endpoint !=null) {
            from(endpoint).to(aggregate)
        }
    }
    from(aggregate).routeId(ROUTEBUILDER_ID)
        .log("Payload ${body}")
        .process(e->threadStopWatch.start())
        .bean(config, "unmarshalObject")
        .bean(mapper, "map")
        .marshal()
        .json(JsonLibrary.Gson)
        .threads(min_size,max_size).id(ROUTEBUILDER_THREAD_ID)
        .process(e-> {
            long elapsedTime = threadStopWatch.getElapsedTime();
            e.getIn().setHeader("elapsedTime",elapsedTime);
        })
        .log("Elapsed time: ${header.elapsedTime} nanoseconds")
        .end();
}

I was spending a lot of time figuring out the Apache Camel inside of a Spring Boot application to create a stopwatch for that specific message. While I thought I did it properly, I'm not certain that each message will get a new ThreadStopWatch and a valid time. I'd like to have a valid ThreadStopWatch for each message.

System.nanoTime is going to give wall time and I realize this is a shortcoming, especially since it's a multi-threaded app. What would be a better timing mechanism less polluted with other threads?