dave_the_nerd
Cadet
- Joined
- Nov 16, 2013
- Messages
- 6
Hello. n00b here, so please use small words.
I built a box last fall - Celeron 1037U, 4GB RAM, 3x 2TB drives in RAID-Z.
I've been getting (and am getting) 60-70MBps transfers, either read or write, (bursts up to 100MBps or so) to any client machine over any NAS protocol I tried. (CIFS, NFS, and AFS.) I'm okay with that performance.
Later on, after checking some threads here, I added an eSATA JBOD (Sil3726 multiplier) enclosure and a Sil3124 card, because I didn't like the thermals in my NAS case. No change in client performance, although ZFS scrubs take longer.
Last fall, I tried creating a file-based iSCSI extent for my Win7 Desktop, which gave me god-awful performance. (Writes were okay, but reads were, like, 5MBps.) At the time, I read a couple forum threads, read how and why ZFS and iSCSI are just crap together, and contented myself with CIFS for a while.
Now, more recently, I added a 3TB HD to the NAS, formatted UFS (not ZFS!) and created another file-based iSCSI extent. I used all the default settings. (512b block, etc.)
AS-SSD Benchmark performance was fine - ~70MBps each way sequential. Write performance was fine - I copied about 1TB of data to it at around 60-70MBps.
But the transfer speed back to the client is still topping out around 5MB/sec. CPU use on the NAS is near-idle. (Except for the ~20% or so that Crashplan is always using.) And it's safe to say that my client machine, network, and HD aren't the bottleneck.
Judging by other posts in the forums, 5MB/sec read from an iSCSI LUN is really bad. So what information do I need in order to identify and address the root cause here? How would I go about getting it?
TIA.
I built a box last fall - Celeron 1037U, 4GB RAM, 3x 2TB drives in RAID-Z.
I've been getting (and am getting) 60-70MBps transfers, either read or write, (bursts up to 100MBps or so) to any client machine over any NAS protocol I tried. (CIFS, NFS, and AFS.) I'm okay with that performance.
Later on, after checking some threads here, I added an eSATA JBOD (Sil3726 multiplier) enclosure and a Sil3124 card, because I didn't like the thermals in my NAS case. No change in client performance, although ZFS scrubs take longer.
Last fall, I tried creating a file-based iSCSI extent for my Win7 Desktop, which gave me god-awful performance. (Writes were okay, but reads were, like, 5MBps.) At the time, I read a couple forum threads, read how and why ZFS and iSCSI are just crap together, and contented myself with CIFS for a while.
Now, more recently, I added a 3TB HD to the NAS, formatted UFS (not ZFS!) and created another file-based iSCSI extent. I used all the default settings. (512b block, etc.)
AS-SSD Benchmark performance was fine - ~70MBps each way sequential. Write performance was fine - I copied about 1TB of data to it at around 60-70MBps.
But the transfer speed back to the client is still topping out around 5MB/sec. CPU use on the NAS is near-idle. (Except for the ~20% or so that Crashplan is always using.) And it's safe to say that my client machine, network, and HD aren't the bottleneck.
Judging by other posts in the forums, 5MB/sec read from an iSCSI LUN is really bad. So what information do I need in order to identify and address the root cause here? How would I go about getting it?
TIA.