BUILD Complete newbie FreeNAS build

[Kosta]

This is a debate akin to 'what religion is better'.

Whatever you prefer is what's best for you.

I personally never have any drives / desktops / servers sleep. They are 100% on all the time, all day, every day. Even laptop's that I may only use once or twice a month are set to 'never sleep'. The only thing I don't have run 24/7 is monitors. And my desktop monitors I manually turn off when I know I'm going away for more than a few hours. I still set windows to never turn them off.

I have some hard drives that are 15 years old, and have been spinning continuously since new. And they still work great. On the other hand, I'm sure there's people that have drives set to sleep after 5 minutes, and end up having 1,000's of power cycles a year and never have problems. It's all up to your preference.

I agree, I didn't want to say which are better or hold longer as it is probably up to a single disk, it was merely a digression.

My only worry are the yearly power-costs if the server has the 100% uptime without HDDs sleeping. Even with HDDs sleeping, the server would be on and use up to 40W of power, which is not a little when it comes to yearly power usage. My personal usage of the server is twice a day, once in the morning and once in the evening, 5-6 days a week, and then probably weekends either working through or maybe wake up couple of times more, as I would set the sleep to 30-60 minutes in general. And then there are those tests that should be ran, or so I've heard...

I personally never had a disk fail on me after years of usage. I had failures directly after buying it, basically didn't work out of the box, I had a WD Raptor which was losing the filesystem (however data was still possible to rescue), but I never had a disk simply fail on me. Probably due to the fact that my disks are kept quite cold through the watercooling boxes installed in my machine. But there would be no WC in the server, so quite another thing.

 

[Kosta]

And in fact, I am now back at square one, as I am thinking of removing the requirement for transcoding on the server. I still have my computer and could do transcoding on it as it seemingly requires lots of power (and it won't get less in the future), and I will be upgrading the computer more frequently than the server. The fact is that probably in general, a Synology or QNAP device will do better in general in terms of powersaving as they actually do standby. If disks fail or not, well, it is very hard to weigh: running 24/7 and expecting no disk to fail for a long time, leveraging the 24/7 power costs vs running standby like I am doing now, cycling HDDs and the server frequently, expecting maybe they fail or maybe not. Will only one HDD fail per year or more or less... There is no answer to this I think. However, when I think that all my data is moving to the server, I will be accessing it more frequently than I do right now, as it holds video files only. And the server will run BT, so downloading for days is probably a frequent thing...
Argh. Decisions decisions.
However, I would like to thank everyone who helped me with this topic, I have my thoughts squared and cleared up now, the decision is upon me ;-)

 

[titan_rw]

I just can't be bothered for the little money that I might save.

I looked up WD Red's power consumption. They idle at about 4w. So assuming you go from 24h a day with 4 drives to 12 hours a day with 4 drives, that works out to a savings of just under 6kwh per month. Around here, that's about $0.50. My entire nas, which was not designed to be energy efficient, is an i5, and 11 7200 rpm drives. It draws about 120 watts, 24/7. Costs me about $7 a month. (These drives use 5.4w spinning) If I reduced the drive spin time to 12h instead of 24, I'd save $1.71 a month. Really couldn't care less. I'd rather just let them spin. That way I know there always there when I need them, and don't have to wait for spinup. I don't have to worry about them spinning up and down unnecessarily when I'm not there (background tasks or such).

I guess if electricity is much more expensive you could be looking at more like $5 a month by spinning down drives. I still wouldn't bother. I'd rather let them run.

 

[Kosta]

Assuming the server uses 80w (approx. number 40w idle with xeon without disks + 4.4w*8) and the 1kWh costs here 0.16c (Eurocents), then that's the yearly cost of €112. That is, for one single appliance, actually really lot, concerning I use it like this:
5 days a week - 1-2 hours in the morning, 4-5 hours in the evening
2 days a week - a whole day, but probably also unusable times of day due to absence
The background processes might take also some time to do, and there is again that mention of torrent running, question is always how long it does that.

The savings here, if I do some simple math, could be up to the half, if not even more, as between 11pm and 6am, there is no real usage really.

While I do welcome the idea of having it up and ready when I need it, and I realize that over year seen, the savings of 50€ might not be much, but 50€ is also not little. Wanna just give me these 50€ per year, like you don't care? Didn't think so :)

 

[cyberjock]

@Kosta

To be completely honest, there's no reason why you can't get something like a g2020 and worry about upgrading later if you need it. It's just a CPU upgrade and FreeNAS won't even care that you've changed CPUs. Heck, even if you wait a year before deciding to do transcoding, you'll probably be able to get the 1230v2 for much cheaper than it is today.

There's 4 main philosophies to building a server(or a desktop):

1. Build it with the absolute cheapest components you can afford. (Typically it seems like when people do this they make way more compromises than they realize and have regrets later)
2. Build it with the components that will make you happy today. Worry about future requirements if/when you come upon them.
3. Build it with more power than you think you will need for the next 12 months or so.
4. Overbuild it with the intention of it lasting so long you won't want to upgrade it later but rather replace the whole thing.

As a modified version of #3, easy upgrades like CPUs and RAM aren't that intrusive and can be done in just a few minutes. Typically I don't recommend people buy more than enough RAM, but don't buy low density sticks to allow for more cost effective upgrades later.

For you, #2 or #3 will probably do just fine.

 

[Kosta]

I really don't care about the buying cost. I will get the CPU I really want. (yes, I am kinda limited by €1500, but I can fit the Xeon in there too).

My problem is the running cost, these are the costs I have to factor in, with a factor of 10 or 15 (that stands for years). I know the technology changes and I will likely have to replace the whole thing one day, but I don't want this day to come in 5 years.

 

[cyberjock]

To be honest, in 5 years you'll almost certainly be forced to upgrade just to support more RAM. There are likely to be 10TB+ drives by then, and your "small" 32GB of RAM won't be able to handle a pool of that size.

 

[Kosta]

Well, I wouldn't upgrade now if the space were enough on my current NAS. I am running out of it, and I filled 2TB in about 3 years. So this is why I wanted 8x2, as that would give me enough for way more than 5 years. Of course depending how the video filesizes also grow, a bit shorter, but certainly longer than 5 years. At least that's my calculation. But of course, that is my calculation, that I want to avoid upgrading components like the case or drives, if not necessary. The key is prolonging the hardware costs and minimizing the power costs. And this is what built vs DIY actually defeats one another. Build = higher hardware costs, DIY = running costs. I reckon both come, at the end, at the same price...