On Jun 2, 2007, at 4:01 PM, Ask Bjørn Hansen wrote: > > On Jun 1, 2007, at 7:32 PM, Matt Simerson wrote: > >>> Our default clock rate is 1,000 (sysctl kern.clockrate). >>> The net effect is that if we're processing less than 1,000 qps, >>> polling is actually more expensive than letting the nic generate >>> interrupts (the default). It's only after that threshold gets >>> crossed that polling delivers an advantage. > > I haven't had a chance to try it, but setting the clockrate to say > 100 or 200 could help on this, no? > > - ask
Almost certainly! The key to predicting whether polling is advantageous or not is firing up your favorite network analysis tool and seeing what type of traffic your system is seeing. If you want to tell retroactively, install something that monitors CPU usage and see what difference it make before and after. On a small network, the type well suited to a net4801, I should think that in the vast majority of cases, a guaranteed 200 polls per second, even when the network is idle, is more load on the system than simply processing the packets as they arrive. YMMV, but even in recent builds of 7.0, the cost of per-packet processing overhead is still quite high compared to 4.x. Matt _______________________________________________ Soekris-tech mailing list [email protected] http://lists.soekris.com/mailman/listinfo/soekris-tech
