Possibly ideal mini-ITX NAS platform

Status
Not open for further replies.

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
No, I think your original gameplan is promising. The problem is you are the first to be daring with the E3... no one else on the forum has experience with it in that chassis. I think there is at least some risk, but it is such a promising option, people here WILL end up doing it.

So that's why I spent some extra time discussing options. There is ABSOLUTELY NO REASON that I see that it can't be made to work. I was not trying to discourage you, but instead offer insight into the potential problems and also how we might have addressed the issue were we building it here in shop. It is very hard to do that without actually laying hands on, but I would feel comfortable ordering such a system ... the likelihood it is "unfixable" strikes me as very low.

Think you should plow ahead, myself. Assuming you have access to basic materials (sheet plexi or plastic) and a Dremel tool, there is nothing I see that is likely to present great difficulty.
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
No, I think your original gameplan is promising. The problem is you are the first to be daring with the E3... no one else on the forum has experience with it in that chassis. I think there is at least some risk, but it is such a promising option, people here WILL end up doing it.

So that's why I spent some extra time discussing options. There is ABSOLUTELY NO REASON that I see that it can't be made to work. I was not trying to discourage you, but instead offer insight into the potential problems and also how we might have addressed the issue were we building it here in shop. It is very hard to do that without actually laying hands on, but I would feel comfortable ordering such a system ... the likelihood it is "unfixable" strikes me as very low.

Think you should plow ahead, myself. Assuming you have access to basic materials (sheet plexi or plastic) and a Dremel tool, there is nothing I see that is likely to present great difficulty.

I don't have the tools nor the inclination really to customize the case, I have little time for this "hobby" of mine as it is right now. That's why I went with an i3 instead of the E3 hoping the heat would be less.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
That's why I went with an i3 instead of the E3 hoping the heat would be less.

Oh, that can be such a tease.

Here's an example:

Your i3-4130 has a TDP of 54w(2c/4t).

My server has an E3-1230v2 and has a TDP of 69w (4c/8t).

Let's play a game called "So how efficient is your CPU?" starring you and me. :)

Let's pretend that we both decided to scrub systems that were identical except for the CPU. Lets also pretend that our CPUs ahve the same benchmarks on a "per core" basis for simplicity. We will use my pool as an example- I have an 18 drive RAID3(don't ask and I don't recommend it). Your CPU would end up at full throttle using all 54w as my pool needs slightly more than 2 cores for a scrub. Your system would be at 100% the whole time, and for longer, since the CPU is the bottleneck. My CPU however, being a more efficient CPU for the workload might perform the scrub faster but still at a lower total wattage. So literally, your CPU will get hotter than mine and your scrub will be longer than mine.

Now you're probably going to argue that your CPU isn't the same as mine on a "per core" basis and you'd be right. But, that muddies the waters even more and can make it more difficult to actually determine which CPU is "better".

The TDP is nothing more than the maximum theoretical heat load your CPU can output. The fact that you have 1/2 the cores I do but only saved 15w tends to show that your CPU is likely much less efficient than mine(not to mention slower). To be honest, I'm not sure what it would take to get my CPU to 100% as I haven't been able to get the darn thing above 50%. But your CPU, I could easily make that sucker break a sweat. :)
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
Note that I'm not saying you made a bad choice with that CPU. I'm simply saying that your assumptions for which CPU will be "cooler" is almost certainly incorrect.
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
Note that I'm not saying you made a bad choice with that CPU. I'm simply saying that your assumptions for which CPU will be "cooler" is almost certainly incorrect.


I appreciate that information I hadn't really considered that. I guess I was just turned off by the E3 for two reasons:

1. The maximum heat output in a tiny case with little to no airflow (which you seem to have debunked).

2. The temptation to do much more with this build than to use it only as a dedicated NUS device.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
I appreciate that information I hadn't really considered that. I guess I was just turned off by the E3 for two reasons:

1. The maximum heat output in a tiny case with little to no airflow (which you seem to have debunked).

2. The temptation to do much more with this build than to use it only as a dedicated NUS device.

For #2, that wouldn't hold me back. If I were wanting to start using plex and needed more processing power I'd be shopping for another CPU. I have plenty of CPUs laying around. I really don't need another one. But in the argument between building a new machine to run Plex on or buying a new CPU, I know which one is cheaper ;)

Also keep in mind that tiny cases with little airflow can(and often do) cook your hard drives, so they don't necessarily make "good" NASes either.

Any other bubbles I haven't burst? Should I kick your dog too?

Sorry to be a debbie downer, but it is a mess with lots of variables, and with each choice there's always a downside. Now you see just a very small tip of the iceberg of the kinds of engineering problems Intel has with distributed heat loading, power saving features, and performance enhancing features on the same CPU.
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
Well here's the thing, I have plans (over the next few weeks) for a newly built server to run Plex and a few other VMs. The whole point of me considering going to ZFS in the first place was to separate my storage from my production servers.

My current home file server (with an 8 drive RAID10 array) is a Windows Server 2008 R2 box that runs Plex and Subsonic as well as does file sharing for my home network. It's no longer powerful enough (Q6600 with 8GB of RAM) to keep up with my media streaming needs and I'm tired of having a bloated Windows Server OS strictly for a file serving purposes. That is what led me to build a new dedicated storage server and what ultimately led me to this forum in hopes that FreeNAS would be my best option for a dedicated storage OS. And since I want this server to strictly protect my data and serve them up to computers and VMs I didn't think I needed huge amount of CPU power. But you're telling me different? Or are you just saying to go with a more powerful CPU for the headroom of being able to do more things in the future with it (something I won't need)?

I realize that the U-Nas box is far from ideal cooling wise. However the decision to go this route is not much of a decision at all but more of a necessity. I don't have the room for rack mountable hardware and while I could POSSIBLY fit another large tower it would go against my original intention of reducing the amount of huge PC towers I have in my office area (just sold my huge powerful gaming PC for this reason).

Therefore I'm sort of left with a real desire and need to go SFF for my storage needs and not exactly sure of the best way to get myself there. I realize that the vets on this board have the attitude that "saving money now just leads to spending money later" and that cheaping out is just playing with fire in terms of your data security. Just keep in mind this is not a cost saving venture, it's a space saving one. I know that I'm not in a position to go after the "ideal" or "optimal" setup for a FreeNAS server so with that in mind I'm looking for the best possible solution given my needs.

I wasn't under the impression that FreeNAS was an all or nothing OS where you either go full bore and do server grade EVERYTHING or you don't go FreeNAS at all. I'm hoping that's not the case but if it is I'll gladly find a more suitable solution for what I'm trying to do.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
See the business part of all this is

http://cdn.overclock.net/d/dc/900x900px-LL-dcdc30c5_DSCN0165.jpeg

which gives you a feel for the relatively tight confines available; near as I can estimate based on

http://cdn.overclock.net/f/fe/900x900px-LL-feb46033_DSCN0164.jpeg

and

http://www.u-nas.com/xcart/images/D/back3.JPG

Now it looks to me like the cavity is at least as tall (wide? since it's sideways) as the I/O shield. This is really good because the I/O shield is itself 1.75" (i.e. the same amount of space as a 1U) so there's actually more space there than there would be in a 1U. That's a very favorable situation in some ways. But it is sort of hard to visualize sizes. I prefer to lay hands on stuff for that reason.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
My current home file server (with an 8 drive RAID10 array) is a Windows Server 2008 R2 box that runs Plex and Subsonic as well as does file sharing for my home network. It's no longer powerful enough (Q6600 with 8GB of RAM) to keep up with my media streaming needs and I'm tired of having a bloated Windows Server OS strictly for a file serving purposes. That is what led me to build a new dedicated storage server and what ultimately led me to this forum in hopes that FreeNAS would be my best option for a dedicated storage OS. And since I want this server to strictly protect my data and serve them up to computers and VMs I didn't think I needed huge amount of CPU power. But you're telling me different? Or are you just saying to go with a more powerful CPU for the headroom of being able to do more things in the future with it (something I won't need)?

Well, it's like this.. either you are CPU bound or you aren't. There really isn't a whole lot of room for "I'm cutting it close" as its difficult to guesstimate your needs(that's something you'll have to decide for yourself). If you want to run FreeNAS for CIFS shares only a g2020 CPU is a great fit. It'll saturate Gb and it'll do scrubs reasonably well. Naturally, if your scrubs are CPU bound with a g2020 its not the end of the world if it takes longer, right? Even if you only got 80MB/sec instead of 95MB/sec during a scrub you won't be too terribly upset about it. But what if your CPU is so underpowered it takes 3+ days to scrub and the whole time the server is less than 10MB/sec without any plugins? Then you'll be pretty upset, and rightly so. I had that situation with my first FreeNAS server until I upgraded my RAM(I had insufficient RAM for my pool size). In my case a scrub would never complete before the next scrub would want to start, even with me not trying to access the CIFS shares. That's just unacceptable.

But as soon as you start doing things that have specific timing needs(streaming transcoded video via Plex, doing iSCSI and/or NFS for ESXi Datastores, etc) you are forcibly requiring a certain amount of CPU power. How much depends on your needs and what VMs will typically be doing, etc. That's only something that you can realistically guesstimate. Guesstimate too low and you WILL be upset when Plex can't stream a movie. Go too overboard on the CPU and you'll have spare horsepower(maybe even too much) and your wallet will notice more. It's a decision that you have to make for yourself based on what you plan to use the server for. Personally, I can't imagine an E3-1230v2 being "underpowered" for anything I'd want to use a FreeNAS box. Even if you installed every plugin we had and ran an ESXi datastore on ZFS its totally doable with that CPU. So if %randomuser% is okay with being fairly sure that they won't be CPU bound, I consider an E3-1230v2 to be an excellent choice. It has ECC RAM support, AES-NI, and is plenty powerful enough for anything I can imagine. In my opinion, its the ultimate CPU to buy if you aren't 100% sure of your future needs because of what it offers. AMD probably has something roughly equal but as I don't keep up with AMDs I won't try to provide bad advice for that.

I realize that the U-Nas box is far from ideal cooling wise. However the decision to go this route is not much of a decision at all but more of a necessity. I don't have the room for rack mountable hardware and while I could POSSIBLY fit another large tower it would go against my original intention of reducing the amount of huge PC towers I have in my office area (just sold my huge powerful gaming PC for this reason).

Its totally your call on what you define as "necessity" or not. Me personally, I'd rank necessity to keep things cool for longevity over size. But that is strictly a personal choice. If you want to build a hotbox with FreeNAS, that is totally your choice. As long as you are happy with any potential consequences then who cares if I wouldn't do your build. It's YOUR build. Our(the more senior posters here) words of warning are because we've seen what people seem to get wrong regularly. And 99% of new users can't seem to grasp some of the basic concepts like "make sure your hardware is compatible with FreeBSD before purchasing it" and "make sure you have enough RAM/CPU power". Those are big pitfalls that many users underestimate because they have no experience with FreeBSD/FreeNAS. Plenty of people will assume we have some great agenda to keep the RAM companies and Supermicro in business. The reality of it is that you will rarely find a cheaper setup than some of the ones we offer. If there was a better set of options we'd probably be recommending those. We don't have any brand loyalty to Supermicro, it just works. Right now people are checking out http://forums.freenas.org/threads/mini-itx-c226-haswell-build.15371/ as it might be a great alternative for many users. My only fear with the miniITX design is people are going to underestimate the cooling needs for their hard drives. Their failure rates go way up when you go at or above 40C. I have found very few cases I'd even try because of their lack of cooling. Most of them I wouldn't pay for until someone could prove they could provide cooling.

Therefore I'm sort of left with a real desire and need to go SFF for my storage needs and not exactly sure of the best way to get myself there. I realize that the vets on this board have the attitude that "saving money now just leads to spending money later" and that cheaping out is just playing with fire in terms of your data security. Just keep in mind this is not a cost saving venture, it's a space saving one. I know that I'm not in a position to go after the "ideal" or "optimal" setup for a FreeNAS server so with that in mind I'm looking for the best possible solution given my needs.

I wasn't under the impression that FreeNAS was an all or nothing OS where you either go full bore and do server grade EVERYTHING or you don't go FreeNAS at all. I'm hoping that's not the case but if it is I'll gladly find a more suitable solution for what I'm trying to do.

It isn't an impression, and it isn't a reality. There's some needs that have to be met for ZFS. It's not so much that FreeNAS needs it as much as ZFS does. If you go UFS you'll find that your RAM requirements shrink to almost nothing. 2GB of RAM is all you need pretty much regardless of your disk size with UFS. But, many plugins won't run on UFS. So there's tradeoffs. There's also a distinct possibility that UFS might be stripped out of FreeNAS along with the x86 going bye-bye a year or more from now.

The game is to understand what you are getting yourself into and what you will need for your system. There's rumors that Windows may have worked on something like ZFS a few years ago but bailed on it because they predicted major support nightmares because your stereotyped Windows minion will NOT heed warnings to add more RAM. They'll assume 8GB of RAM is plenty and they create support tickets when it isn't enough and cry that its slow and demand a fix. And Microsoft spends alot of money on support, so supporting things that might cost them significant support resources is something they must consider. I don't know how true/untrue it is because I don't work with Microsoft(thank god?).

Also notworthy is this... You are not building a desktop that, if it crashes or breaks, you simply reinstall your favorite OS and keep going. You are building a server and entrusting it with your data. "if it crashes or breaks" isn't something you should be okay with under any circumstances. Imagine if you had to wipe your FreeNAS server as often as you have to reinstall Windows. You'd never trust your server again. That stability comes at a price. You might save $1000 on a license of Window Server, but you will spend some money getting appropriate hardware(hardware that you should have probably bought with Windows Server but probably didn't). I wasn't building my servers with server grade parts until last year. At some point you'll have to start rolling with the big boys, which means money will need to be spent. :(
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
I wish my inability to build in a tower/rack was due to limited resources because if that were the case I could just wait and save money. This is a space issue and I know I keep bringing it up but I do so because I feel like no one takes it seriously when I say it. I do not have the space for a tower anymore after my girl moved in with me so for the time being lets just forget that tower builds and rack builds even exist. If you're telling me that knowing that fact, I shouldn't go ZFS then that's fine, that's what I'm looking to hear (honesty that is). If you're telling me it's not ideal but doable then that's what I'm looking to hear as well as how to make the best out of my circumstances. So at this point it's either:

1. Yes there is still value in going with a SFF FreeNAS build (meaning I won't see a huge drop off from my current file server performance)
2. Going SFF just wont' allow me to get anywhere near the performance of my current RAID10 setup and it makes little sense to go with a FreeNAS build

If 1 is the answer that's why I'm here, to determine how to make the best SFF build I can. If 2 is the answer, I'm quite willing to stick my Perc 6i in this U-Nas box with a cheaper board and 8GB of non-ECC RAM and be on my way.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
We can't answer your questions. You asking us to tell you what kind of car to buy. It doesn't work like that. You have to decide for yourself what you want and are willing to spend money on. We've given about all the advice we can give. Its up to you to apply that advice for your situation.
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
We can't answer your questions. You asking us to tell you what kind of car to buy. It doesn't work like that. You have to decide for yourself what you want and are willing to spend money on. We've given about all the advice we can give. Its up to you to apply that advice for your situation.

I'm not looking for someone to tell me "yes go with FreeNAS" or "no don't do it unless you go full bore on the hardware." What I've been looking for is some advice on what I can expect performance wise out of a SFF build like the one I've been talking about in here. I think I'll take my questions to the_tox's thread though since he seems to be the only person in here who has recently build something along the lines of what I'm looking for and can give me some insight.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
There is nothing particularly magic about a "SFF build", except that the heat dissipation challenges are greater. If you can cram parts that closely resemble recommended parts in terms of function into a mini-ITX, it will be nearly indistinguishable whether you've built in a monster tower or a SFF chassis, and will perform the same in both cases.

Not having actually laid hands on the hardware you propose, it is very difficult to do anything more than to give you some general guidance. The killer problem is getting rid of heat (or possibly not making it to begin with). cyberjock gave some reasonable advice, more detailed than I would have.

But as an example of what we measured here on the bench with a 12-core E5 system:

Idle the system took 225W.
One core pegged took 260W.
Four cores pegged took 285W.
12 cores pegged took 350W.

So going from 4 cores and spinning up an extra 8 takes 65 more watts, or approximately half the TDP of the CPU (130W).

The first core is the painful one, you add 35 watts to spin it up. As you go with more cores, they get progressively less power-hungry (but also less turbo). Now, each CPU is a bit different, so I can only generally say that a quad core CPU that isn't being fully utilized will be dissipating less heat than one that is. Where that actually lands you in terms of practical dissipation requirements is hard to predict. I would tend to favor a faster CPU with more cores than I needed as long as I could assure that I could meet the required dissipation under unanticipated full load. One could even cheat by cutting the active cores or something like that, but ideally what you want is a box that is able to leisurely go about its expected workload, and a scrub, and a replication, simultaneously, without capping out its CPU - all while being able to handle the dissipation required at full CPU load.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
See http://forums.freenas.org/threads/best-use-of-available-drive-space-for-freenas.15611/

From that post:

Rant: I'm done trying to create a shoebox-sized server; in the long run I've spent more $$$ trying to do that than if I'd of just heeded the advice not to to begin with, just trying to troubleshoot network issues and random file corruption...but that's a story for another whiskey... (Luckily, I was smart enough to keep backups of all these files on external drives.)

We really aren't talking out our butts. We see people do stuff and we see what does and doesn't work, which is where we base our comments to newbies like yourself. We don't have to build the system to see what's wrong(which you said when you said you were going to go to "the_tox's" thread).

There is a tradeoff with everything. Do you put 10W30 or 5W20 in your engine? Different people have different opinions and different expectations. That's why I've said twice(and now three times) that you are the best artist for your server as you are the only one that can make that choice. And we also try not to tell you that it will work for certain because if something doesn't work, overheats, or generally doesn't perform plenty of posters have come here asking for us to send them a check for new hardware because they made poor choices.

The decisions are yours to make. Not ours, nor can we try to infer what is or isn't acceptable risk for you.
 

JimPhreak

Contributor
Joined
Sep 28, 2013
Messages
132
Last thing I'm debating is whether or not to keep my i3-4130 (54w) or trade it for a i3-4130T (35w) to keep internal case temps down.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
what are your temps now?
 
Status
Not open for further replies.
Top