On Fri, May 10, 2013 at 01:12:19PM +1000, yudi v wrote: > I am using the below command/s to write to 3TB hitachi deskstar 7200RPM > disks. > > dcfldd if=/dev/urandom status=off | PV -s 3000G | dcfldd of=/dev/sda > of=/dev/sda of=/dev/sdb status=off bs=4096 > > on the HP microserver nl40 with SATA 3Gbps connection, the throughput was > 3-4MB/s, around 10.5 days. > Where as on my desktop with i5 CPU and 6Gbps connection, the throughput is > 13-14MB/s, around 61 hours. > > My question is: where is the bottleneck on the HP for such low throughput? You could check to see if it's the RNG by doing dcfldd if=/dev/urandom status=off | pv -s 3000G | dcfldd of=/dev/null \ status=off bs=4096 And you could check it's the disk by doing similar but reading from /dev/zero. Also, I don't know if it's a typo here, but do you realise you've specified /dev/sda twice in your output command? That could be an issue. The other alternative is to use a bigger block size. The suggestion here[1] is to set bs equal to your hard drive's buffer size. [1] http://forums.freebsd.org/showthread.php?t=11585
Attachment:
signature.asc
Description: Digital signature