On Fri, Mar 28, 2014 at 6:06 PM, Jorgen Lundman <[email protected]> wrote:

>
>> I don't know enough about these tests to make any real definitive
>> statements, but the results show HFS peaking around 2GB/sec, and ZFS
>> peaking around 6GB/s. Sounds good to me, though it's perplexing that
>> throughput goes down as file size goes up. --matt
>>
>>
>>
>>  Yes, I'm not entirely sure what is going on, but basically just using
> time to check how long a iozone -A takes to run:
>
> hfs# time iozone -A
> real    6m52.913s
>
> zfs# time iozone -A
> real    17m2.767s
>
>
> Pulling out the end data, randomly:
>         Output is in Kbytes/sec
>               KB  reclen   write rewrite    read    reread
>
> hfs:
>           524288       4   50259   57859   498526   677862
>           524288       8   69938   72726   596500   888537
>           524288      16   70807   77607   778664   942680
>           524288      32   80059   64811  1131027  1142847
>           524288      64   64146   76311  1150897  1135341
>           524288     128   64722   76543  1188676  1220951
>           524288     256   61233   64857  1234472  1272219
>           524288     512   59642   66703  1216507  1196796
>           524288    1024   64701   57436  1095813  1061215
>           524288    2048   89257   71440  1011559   997499
>           524288    4096   59390   62918   816415   814473
>           524288    8192   70201   61838   807098   816163
>           524288   16384   55456   62131   770388   675465
>
> zfs:
>           524288       4   69946   61442    35812    33750
>           524288       8   79214   59965    35276    34657
>           524288      16   79565   61427    37954    37884
>           524288      32   78144   58469    35089    39305
>           524288      64   79383   62301    33978    38365
>           524288     128   78349   63518    35415    37586
>           524288     256   79343   64975    38946    35096
>           524288     512   76837   65248    36305    37858
>           524288    1024   80828   65693    35733    36306
>           524288    2048   78684   64832    35045    34183
>           524288    4096   74112   60109    37463    35811
>           524288    8192   78958   61731    31899    45881
>           524288   16384   77999   61969    40890    44041
>
> Which would imply my read and reread values are pretty poor. flatlined
> around 30MB/s, when hfs is over 1GB/s.


Is it possible that the reads from HFS are cached in memory, and the ZFS
reads are coming from disk?

--matt


>
>
>
>
> _______________________________________________
> developer mailing list
> [email protected]
> http://lists.open-zfs.org/mailman/listinfo/developer
>
_______________________________________________
developer mailing list
[email protected]
http://lists.open-zfs.org/mailman/listinfo/developer

Reply via email to