New build: benchmarks and reviews...

[jkh]

I can see exactly where you are coming from, and I'm guessing you work for ixsystem or a company providing similar type of product/services: enterprise class storage systems.

Full disclaimer: Yes, I do work for iXsystems!

The test I made were mostly for my own and typical immediate use: few users, NFS and windows clients. I won't use iscsci at home, nor will I use virtual environment: the projects I'm currently working on do not deal well without direct hardware access (and that includes mythtv and DVB/ATSC acquisition cards). I doubt anyone using freenas at home is going to care much about how it will behave with 500 simultaneous connections, and home users seem to be the greater percentage of freenas users.

Well, this is where things get kind of... complicated.

First off, really pushing filesystem benchmarks for a home user scenario is, as Josh indicated, already kind of pointless since once you've saturated a 1GbE interface all the performance in the world on your filesystem frankly won't matter since you can't GET to it at those data rates anyway. Most home users also just want to store their movies, music, backups, and other data which spends most of its time at rest, so again, if one particular kernel gives them 14% better theoretical throughput numbers that they'll never actually see any tangible benefit from, they just don't care! They just want that storage solution to be reliable and work when they need it, and most of their concerns (as Josh also pointed out) will be around "ease of use", not "raw blazing speed" (this is also why mom does not buy a Corvette Z06 to take the kids shopping in). That is an entirely different area of engineering endeavor, and one we actually value quite a bit since not just home users want a box to be easy to set up and use, so do many enterprise users, actually. Even the most savvy enterprise person has better things to do during the day, generally speaking, than geek out on some highly arcane interface that seems to enjoy making what should be fairly routine and simple tasks complex!

Second, and the reason "real world performance" can't simply be ignored in favor of ease of use, is that home users frequently turn into enterprise users, just as iPhones and iPods have been "invading" the workplace in record numbers. Something that works well at home becomes tempting to use at work, and then that very same software solution gets weighed on a different set of scales because at work, they DO have multiple 10GbE interfaces and hardware that costs 5-6 figures and can easily serve hundreds or even thousands of users. It's a tough line to walk, since the very same software is used in both scenarios - you don't get the luxury of focusing on just one narrow patch of the market.

 

[jyavenard]

I'd be very interested in knowing what sysctls and tunables you had set as well as the output of zpool status.

it was absolutely stock config in both cases. The only tunables I added to freenas is loading the ipmi kernel so I can read the temperatures and fans speed.

We also have a number of big performance improvements just around the corner in FreeNAS 9.2.0. Would you be interested in a sneak peek at that?

can't see why not.. is it on github?
(btw, right now compilation fails with: "=> SHA256 Checksum mismatch for libgpg-error-1.11.tar.bz2. [snip] ===> Giving up on fetching files: libgpg-error-1.11.tar.bz2 "

To answer Jordan's point about performance under scale, it's fairly easy these days to build a storage rig that will saturate gigE. Unless you have multiple clients or are using a block technology that can do MPIO, or can afford 10Gbe you're limited by the performance of gigE. In that case it doesn't make much difference if your volume can go 200 or 2000 MB/sec.

If you are in a bigger scenario, where you have multiple clients or can do MPIO or have 10Gbe, or some combination of the above, then you almost certainly are concerned about scalability, in which case Jordan's observations do apply.

well, that's the thing... I can do > 300MB/s read or write locally. But I'm far from maxing out the gigabit connection.
with NFS I can get peak of 120+MB/s over NFS; with smb I very rarely speed over 100MB/s

Now, smbd never goes over 25% of CPU here (and no, that's no 100% on one core, 25% on one core). I tried fiddling with the buffer size in smb.conf without much success.
On FreeBSD, I remember having to enable the aio kernel module and enable aio support in samba which made a significant speed impact. But on freenas I can't tell if it's loaded or compiled statically or if samba makes use of it..

What is certain, is that I wonder what folks like Thecus or Qnap do to systematically maxout the gigabit link over windows share using the Intel NASPT when they use 2gig of ram and an atom processor !

Must be some wizardry in there, because with better disk controller, more ram, better processor I only manage half that ! (and that's either using linux or freebsd)

However, please don't be concerned too about our focus. I just spent many hours last week dialing in mDNS and updating FreeNAS to a new version of netatalk (AFP). We do care about home users. In fact we have a big GUI revamp coming down the pipe in 2014 just to address the types of users that just want the damn thing to work with no fuss or muss

this is great to know !
While I consider myself a power user, not being able to recommend such and such product because it's either easy to use but give shocking performance (read dlink, linksys) or it's powerful but it's so complicated that I know as soon as they are going to want to do something they will give me a call.

Don't get me wrong.. FreeNAS from what I've seen and tried is already the most user friendly zfs nas frontend available, it's just that there's a long way to go before user X can use it without knowing what a vdev or dataset it.

What I need to salute you guys for is the quality of the manual and instructions... It's really really good. Everything is there for those seeking for it...

Talking about netatalk; I've noticed that if I add a share ; none of my macs will now be able to connect (always an error about either the share having disappeared or the password is incorrect) until I go back into freenas; disable AFP and turn it back on again...
Also, does the AFP password has to be limited to 8 characters only? I don't recall that limitation in afp; yet freenas mentions it in the help for the password.

Thanks again for tuning in...

 

[jpaetzel]

We'll put an image up to test tomorrow.

 

[jyavenard]

I've given up trying to control the fans... The SM Bus kernel driver is just too old on FreeBSD is several hardware revision behind everything.
So I got this instead:
http://www.camera2000.com/en/speed-controller-by-pwm-for-computer-pc-cooling-fan-4pin-3pin.html

This is a little voltage adapter that works in addition to PWM, that will allow the fan to run at a % of what the motherboard set...

 

[jyavenard]

We'll put an image up to test tomorrow.

any update on this? where would it be available?

 

[Johhhn]

Next time this happens, before disabling AFP and re-enabling, can you try connecting to it via the Go to menu?

Even when I can't connect like you mentioned below, it still works when I connect to it 'manually'.

Talking about netatalk; I've noticed that if I add a share ; none of my macs will now be able to connect (always an error about either the share having disappeared or the password is incorrect) until I go back into freenas; disable AFP and turn it back on again...
Also, does the AFP password has to be limited to 8 characters only? I don't recall that limitation in afp; yet freenas mentions it in the help for the password.
Thanks again for tuning in...

 

[jyavenard]

I only ever connect using the Go To menu. I find that using the finder network view takes way too long

 

[Johhhn]

Ahh, sorry,, I mis-read what you wrote. I've seen the same behavior occur as well.