Why does this warning comes up?

[lohmag]

With CIFS i can utilize all space without problems?

 

[pirateghost]

No matter what option you use, you should be wary of filling your pool over 80%.

Please read the documentation and cyberjocks zfs intro power point.

 

[jgreco]

Typically backups will cause a lot of fragmentation on a ZFS pool too, though it will probably be a lot more favorable for file-based storage.

 

[depasseg]

My guess is the concern with cifs is the transfer speed being limited by the single threaded performance of a CPU.

 

[Nick2253]

Samba being single threaded has been completely blown out of proportions.

First off, Samba is only single-threaded per connection. So, if you have ten different people connecting, each connection will be its own thread. You're not bottlenecking everything through one core.

Second, only the data transfer component is part of Samba. The actual data processing (retrieval, writing de/encrypting, etc) is part of FreeNAS, and will use other threads. For any modern processor, including an Avoton, the single threaded Samba performance is more than enough to saturate a 1Gb ethernet link. With unlimited network bandwidth, a modern Pentium would be more than capable of serving the max speed of a typical 6-disk RAIDZ2 array, in which case your array, not your processor, would be your limiting factor.

And lastly, even if you got to a place where your single-threaded Samba performance was your bottleneck, I'd wager that your CPU is now seriously undersized for the rest of your system, and you'll have bigger problems than Samba.

 

[Ericloewe]

Hell, the puny Silvermont Atom core in Avoton can push a 10GbE connection past 5Gb/s. CPU performance for CIFS really isn't a realistic issue in most use cases.

 

[jgreco]

Hell, the puny Silvermont Atom core in Avoton can push a 10GbE connection past 5Gb/s. CPU performance for CIFS really isn't a realistic issue in most use cases.

I can't imagine it'd be doing anything super-useful while doing that though. Running iperf, yes, I guess that doesn't shock me. We've been able to peak out a 1Gbps connection with a single CPU core since the Pentium 3 or 4 days...

 

[anodos]

I can't imagine it'd be doing anything super-useful while doing that though. Running iperf, yes, I guess that doesn't shock me. We've been able to peak out a 1Gbps connection with a single CPU core since the Pentium 3 or 4 days...

I think I managed to saturate gigabit (or came close to it) once with samba on a WD MyCloud.

 

[jgreco]

I think I managed to saturate gigabit (or came close to it) once with samba on a WD MyCloud.

That was actually one of the things that initially made me so unhappy with ZFS. Our 2005-era storage platforms (Opteron 240EE) were fully capable of saturating gigE's when used with FreeBSD and UFS, but when I put FreeNAS on them with ZFS, it was like going back to horse and buggy days.

 

[Ericloewe]

I can't imagine it'd be doing anything super-useful while doing that though. Running iperf, yes, I guess that doesn't shock me. We've been able to peak out a 1Gbps connection with a single CPU core since the Pentium 3 or 4 days...

That's what the other 7 cores are for. :D