[bind10-dev] recent performance improvements

Jeremy C. Reed jreed at isc.org
Wed Mar 28 23:16:02 UTC 2012


> I also have macmini results versus BIND 9.9.0, but don't consider them 
> too accurate. I need to make sure no other heavy services or jobs are 
> going on at the same time.
> 
> root-memory-.nxdomain 34416.501217 qps 14% slower
> root-memory-.soa 17150.547136 qps 26% slower
> root-memory-.success 17204.327813 qps 16% faster
> root-memory-cpus4.nxdomain 25088.150922 qps 25% slower
>   but another time it was 2% faster and  32792.762246 qps
> root-memory-cpus4.soa 21092.460988 qps 52% faster
> root-memory-cpus4.success 21830.332103 qps 57% faster
> smallzone-memory-.nxdomain 34267.082050 qps 19% slower
> smallzone-memory-.soa 34209.730366 qps 10% slower
> smallzone-memory-.success 34469.664609 qps 12% slower
> smallzone-memory-cpus4.nxdomain 33240.307671 qps 7% slower
> smallzone-memory-cpus4.soa 31214.336175 qps 0% faster
> smallzone-memory-cpus4.success 33798.697208 qps 2% faster

Notice that the Linux four b10-auth results are around twice as fast as 
a single b10-auth. But on the above MacOSX, the four b10-auth runs are 
slower or at least not significantly faster. This system has:

$ sysctl hw.physicalcpu  hw.logicalcpu        
hw.physicalcpu: 2
hw.logicalcpu: 2

I will provide results for other amount of processes later.







More information about the bind10-dev mailing list