Grizzly RoundRobinConnectionDistributor 的变化

grizzly1.9:

private SelectorRunner getSelectorRunner(int interestOps) {           SelectorRunner[] runners = getTransportSelectorRunners();           int index;           if (interestOps == SelectionKey.OP_ACCEPT || runners.length == 1) {               index = 0;           } else {               index = (counter.incrementAndGet() % (runners.length - 1)) + 1;           }                     return runners[index];       }   

一个SelectorRunner 负责accept,其他负责read/write

grizzly2后

private SelectorRunner getSelectorRunner() {          final SelectorRunner[] runners = getTransportSelectorRunners();          final int index = counter.getAndIncrement() % runners.length;                    return runners[index];      }   

不再区分accept/read/write 的SelectRunner.

咨询了作者,主要的架构权衡的问题，如果是系统的连接不多，如果区分accept和read/write,accept的 SelectRunner就会空闲，无事可做，而read/write 的SelectRunner很忙。如果连接很多，可以考虑类似1.9的实现，区分accept和read/write的 SelectRunnerSelectRunner.