Initial impressions of DFly

Kris Kennaway kkenn at xor.obsecurity.org
Fri Jul 2 13:29:00 PDT 2004


On 2004-07-02, Jonathan Fosburgh <syjef at xxxxxxxxxxxxxx> wrote:
>
> On Thursday 01 July 2004 04:23, you wrote:
>
>>
>> Try comparing the performance of tarring and untarring very large files
>> on your hd. What is your subjective impression of DF performance on this
>> test vs FreeBSD 5.x and 4.x?
>
>
> - From 5-CURRENT updated a couple of weeks ago, extracting www.firefox:
>
> make extract  68.98s user 21.59s system 55% cpu 2:43.39 total
>
> And from DFly 1.0RC1:
>
> make extract  67.03s user 21.26s system 61% cpu 2:23.36 total
>
> Now this was done on an otherwise quiet system.  On DFly, running KDE with 
> konqueror, kontact, and several other apps open and while listening to music 
> I received these numbers:
>
> make extract  67.75s user 24.19s system 42% cpu 3:33.93 total.  It was almost 
> unnoticeable, save for a couple of skips in my music. Under 5-CURRENT this 
> extract is somewhat noticeable, though for both OSes it is much more 
> noticeable on builds, though DFly performs better, subjectively, versus 
> - -CURRENT (with ULE).  Even SCHED_BSD felt a little slower, however.

ULE is fairly broken in a lot of workloads and should not be used for
performance comparisons unless you're also going to benchmark
SCHED_4BSD.

Kris





More information about the Kernel mailing list