• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Soft-RAID Read Performance

Z

zerojunkie

n00b
Joined
Mar 2, 2009
Messages
2
I ran across a curious disparity between mdraid and zfs recently I'm wondering if the folks here could shed some light on. I put freebsd 9.0 on a supermicro board with a couple Intel 82573 NICs and three WD 5400RPM drives in a raidz and shared the volume with istgt. Using the Win7 initiator I get about 30 to 40MB/s of throughput from a 7gb file I'm copying down to the client. I thought I'd give some other setups a chance since I already have a similar setup (this box is meant to be a backup of that one, eventually). So, I put Debian stable on the same hardware with MD-RAID and LVM and, to my surprise, got roughly twice the throughput. I know windows explorer isn't the best benchmark out there, but it's the closest test to what I actually do with the data on the NAS. Am I missing something? I figured zfs would be a bit slower, but not this much.
 
You did not compare ZFS vs mdraid. To do this, you may need an iscsi implementation that can work on either. If you like ZFS and look for the probably fastest iSCSI stack with ZFS, you should compare a Solaris platform like the free OmniOS or OpenIndiana with Comstar .
 
A pool with a single vdev in raidZ will get you the speed of one drive. In MDRAID you can get better results. At 40MB/s however something else is limiting your speeds.
 
Aesma said:
A pool with a single vdev in raidZ will get you the speed of one drive. In MDRAID you can get better results. At 40MB/s however something else is limiting your speeds.
Click to expand...

I really doubt.
This is more a general striped Raid item.
If you can read sequential data on Raid-Z1 or Raid-5, the transfer rate can be
theoretically up to rate of a single disk x number of data disks.

If you read small random files, all heads on all disks must be re-positoned for every read.
So the overall (I/O) performance is in both cases like one disk.

The only special with ZFS is, that you can have a lot of striped Raid-vdevs in a pool,
each improves sequential and random performance. (Similar to Raid 50/60 but with more Raids than two)
 
_Gea said:
You did not compare ZFS vs mdraid. To do this, you may need an iscsi implementation that can work on either. If you like ZFS and look for the probably fastest iSCSI stack with ZFS, you should compare a Solaris platform like the free OmniOS or OpenIndiana with Comstar .
Click to expand...

Thanks for the reply. It's funny you should mention that. I loaded up OmniOS yesterday and have comstar setup, but I'm not seeing any better performance than with FreeBSD/istgt. I'm in a throw-down with it now trying to get iperf compiled. Though when BSD9 was on it, iperf reported ~950mbps between it and my client. I'm assuming there's not a network issue here either since I'm seeing nearly the same performance numbers.
 
You must log in or register to reply here.
Back
Top