BUILD My next FreeNAS build: reliable, expandable and easily maintainable.

[joeschmuck]

Damn, I'm going against the grain here but with the costs dropping on the SSD's, I'd recommend finding the cheapest one you can buy if you have the available SATA port. The size of the SSD isn't important as long as it's 16GB or more. If the cheapest one is 120GB, that is fine, grab it. If you don't have a SATA port then both the SSD and DOM is out of the question and I wouldn't buy a DOM anyway, they cost too much. If you must go with a USB Flash Drive, buy a good one, I like the Sandisk Ultra Fit as well, I have a few of those but never used them for FreeNAS. Updates do take longer using a USB device but honestly, how frequently are updates done (LOL, not that frequent now but there are times when something new is released, lots of bug fixes mean lots up updates). I still use my USB Flash device but once my hard drives start to fail, I'm buying fewer larger capacity drives and opening up a SATA port for a boot device.

Also, mirrored boot devices, I'm not in favor of them in general because they do not work as intended because the main boot device has to fail in such a way that it doesn't hang the system and guess what, that is a rare situation, just do some searching in the forums. Make a backup of your configuration file after you make any changes to your system and you will be able to rebuild your system boot device without issue.

And just so every one knows, I do value Dr. KK's advice, he's a valuable source of information, just I don't always agree with everyone on everything.

 

[Pheran]

And just so every one knows, I do value Dr. KK's advice, he's a valuable source of information, just I don't always agree with everyone on everything.

There's certainly nothing wrong with using an SSD or DOM as the boot device, it's just a question of value vs cost. People will make different judgements about the value of a fast boot device to them, so of course we won't always reach the same conclusions. As long as you understand what benefits you are getting from the money you are spending, it's all good.

 

[DrKK]

I'm not at all against anyone using SSD/DOM for their boot device. If you've got 50 euro laying around, and want the extra performance for the 30 seconds per week where it's relevant, then I certainly don't object. I just think it's a solution in need of a problem, for the typical home user. Any "reliability" arguments fall flat to me for two reasons: 1) I haven't had reliability problems, and with a mirrored pair, I have a line of defense in any case, and, 2) Who cares. Let's say the whole boot seizes up. Big deal. If my boot pool dies, I'll throw it in the garbage can, make a new one, and my data pool is right where I left it.

As @joeschmuck says, the failure modes on the USB drives may be inconvenient, i.e., you might hang your system for a time. Fine. If you are doing a commercial deployment, or physical access is a problem, or as close to 24/7 service is imperative, then sure, spend the extra money. But for Joe Blow, the guy with a NAS in the family room, (which is probably 85% of the people reading this), I think the small risk and performance hit he takes by booting from USB is worth the $50 he can now spend on, say, tipping @cyberjock 's mom at the club.

As usual, people will get multiple opinions from people who know what they're talking about, but approach problems with different sets of assumptions.

 

[xdma]

And in any case, your boot-device(s) in no way affects/harms your pool. Your data pool is a separate entity. Should you drop your boot pool, simply replacing it with a fresh boot device(s) is all you need, and you can re-import your existing data pool.

I’m glad to have a confirmation on that.

Regarding the SSD vs. USB pair… It seems that I started a discussion I didn’t want to. Sorry.

I really value all of your opinions (I wouldn’t be here otherwise). Based on them, my own experience and the component prices here in my country, I come to these conclusions.

A pair of Sandisk Cruzer Fit 16GB is 18€
A pair of Sandisk Crizer Fit Ultra 16GB is 20€
The cheapest 32GB SSD is 30€ (I have a choice between Transcnd, Kingston, Sandisk and Adata).

I do NOT need fast updates and I surely do not need 24/7 availability.
On the other hand, I DO have enough SATA ports also for future expansion.

Hence I will go for an SSD. I think that IN THIS CASE it is worth paying 10€ for the added bonuses.

 

[joeschmuck]

Regarding the SSD vs. USB pair… It seems that I started a discussion I didn’t want to. Sorry.

Don't worry, this comes up often and it really comes down to available SATA ports and cost. Everyone will have an opinion and we do like to give them, and we like to provide facts as well and I think we all have done that for you. It will be up to the end user to choose what they desire. And of course I'm glad you went with a SSD, unless you have one die from infant mortality, I think you will be real pleased with the results.

One thing, a SSD does not absolve you from making a backup of you configuration file. Whenever you make any changes to your configuration, you should also save a backup of that configuration in case you have some issue where you must rebuild and restore your configuration.

I hope things work out well and of course, if you have questions, do a Google search first and if that doesn't help, ask in the forums.

 

[xdma]

I have done some calculations regarding the power consumption of the system listed in the first post.
I gathered some information that were not easy to find, so I thought I’d share.

Basically, everything is summarized in the table below.

Here are some comments on my findings.

• All the data are taken from the manufacturer data-sheets.

• The “peak power” is the maximum power that the device can draw as specified by the manufacturer. It is the value that a design engineer should consider when designing the voltage and current regulators. Even though sometimes this value is not representative of a realistic situation (in other words the device doesn't draw so much power under normal use), theoretically the device can draw the peak power under rare circumstances, and that’s why the source should be designed accordingly.

• Sometimes the manufacturer specifies the power that the device needs under normal usage conditions. That’s indicated in the “nominal” column. It can be listed as average in some data-sheets.

• The “peak on 12V rail” column indicates the peak current pulled from the 12V rail. It corresponds exactly to the peak power only if the device is fed exclusively by the 12V.

• The “idle” column specifies the power request when the device is not in use, in idle state, or it works at minimum capacity.

• As it seems impossible to get the power consumption of the motherboard only, I considered the power request of the main components present on the motherboard. What is missing are some minor components like BIOS and Nuvoton, as well as the efficiency of the on-board voltage regulators. Nonetheless the numbers should give a good figure of what is the motherboard power requirement.

• Note that the AST2400 data sheet doesn't report an idle power consumption. So maybe it doesn’t idle, after all it never stops monitoring the sensors and waiting for inputs (it can only clock down the graphic core).

• The data sheet of the LSI SAS9207-8i doesn’t specify an idle power either. On the other hand, one can safely assume that the power drops to zero if disabled from BIOS.

• The power requirement for the CPU seems high, but it’s what Intel specifies for a 4th generation core CPU with 53W TDP. The activated features (like VT-d, AES-I etc.) do not make difference for Intel, so they won’t for me. Some people consider the TDP as a good reference, but that is simply wrong: the design requirements are clear. And that power is drown from the 12V rail (the 8pin additional connector that feeds the VRM of the CPU has only 12V), so it counts entirely for the current evaluation on the 12V rail.

• The fan of the boxed CPU was another surprise. It really needs all that current, more than the hard drives!

Conclusions

• The first thing that I noticed is the huge difference between the idle and the peak power. It is basically impossible to pretend that the PSU will work in its high efficiency bandwidth.

• The G-450 seems exactly right. Actually it is too right as it doesn’t leave any room for expansion. Even with only two additional hard drives, I will be over the 80% of the PSU capacity. BTW I second the good practice to use a PSU to no more than 80% of its capacity.

• The constraint here seems to be the maximum power (rather than the current on 12V), but the situation could be different for another PSU, or if a Xeon instead of a Pentium is used.

• Spinning down the HDDs will decrease the idle power even further, and the PSU efficiency will drop considerably. I wonder if there is any gain in the electricity bill at the end of the day. It would be possible to evaluate that, but I did’t care as I won’t let the drive spinning down.

• One can argue that the peak power is overestimated for a real usage scenario. And that could even be true, but who am I to shamelessly ignore what the manufacturer specified? The system is designed to absorb under some (peculiar) circumstances that power and I won’t put the system in jeopardy just because it won’t reach that peak value under *my own* usage pattern. From an engineering point of view that is plain idiocy.

Hopefully this can be useful to someone else.

 

[Bidule0hm]

You are 100 % right, thanks for this very high quality post ;)