We have a number of branch offices where having an AD/DNS server would make a big difference to user experience. The following characteristics apply:

  • There is no air conditioned server room, and no budget to set one up
  • Offices are located in developing world, so high temperature, humidity and dust are common
  • There are no IT staff in these offices
  • Getting replacement parts generally means sending something from Europe
  • Small form factor is preferred due to logistical challenges of transporting equipment to the office
  • Offices have 2-25 users

I've been thinking about the possibility of deploying low powered flash based mini PCs running Windows 2008 R2 as Read Only Domain Controllers. I see it is possible to get Mini ITX devices with Atom 1.6GHz processor, 2Gb of RAM and mSATA flash based storage (32Gb) (e.g. Soekris net6501-70) This looks like it will meet the minimum requirements for running Windows Server 2008 R2.

  • Has anyone rolled out a production system like this and able to share experience?
  • Am I wrong to think that fanless/solid-state drives are better given the environmental conditions?
  • Would 32Gb storage be sufficient for running RODC for our size of domain (800 users, Group Policy under review)?

5 Answers 5


Valid idea to a common problem we all face at some point. But have you consider a solution like the HP ProLiant MicroServer, it seems to be aimed at the type of work environment you are describing and can be obtained for around $350.00 USD.

Since the ProLiant MicroServer uses a low power CPU heat shouldn't be that much of a problem, small form factor and low cost are equally good.

I guess there is no right or wrong answer.

  • I'm going with one of these, since they are rather cheap at the moment!
    – dunxd
    Mar 29, 2012 at 10:23
  • Update - I've now deployed four of these with another 11 to go. Very happy with this choice. The fanless stuff looked cool, but the cost of the microservers was so low it was hard to argue with.
    – dunxd
    Apr 30, 2013 at 11:16

Yes. I have deployed several Server 2008 R2 boxes on the Supermicro X7SPE-HF-D525 board (64-bit dual-core Atom D525, 1.8 GHz). With 4GB of RAM and two 2.5" SATA drives setup in RAID1 using the onboard ICH9R controller, it prices out well under US$500 and consumes exactly 25 watts of power: i.e., no special cooling requirements. These environments generally have had about 10 users and have used these servers for both file shares and domain services without any performance issues. I have no doubt that this would scale to 20 users without issues, especially if it's being used only for domain services.

Put it in the CSE-503-200B case and it will have exactly the same form factor as a rackmount switch or firewall, so you can mount in a 2-post telco/switch rack:

If the site doesn't have a rack, this server form factor is perfect for mounting flush against a wall. There are some holes pre-drilled in the bottom of the case that make it easy to screw the server onto the wall with the top cover off, and then reinstall the top cover.

Be sure to order the MCP-220-00044-0N bracket if you plan to mount two 2.5" drives for a RAID1 array. Alternately, if you prefer, you can mount drives to the inside of the case using 3M Command adhesive strips. This is especially helpful if you want to install a PCIe card in the server without clearance issues.

Of course, it would perform even better with solid state storage instead of mechanical disks.

  • 1
    I would not reccomend the adhesive strips- tried that once and had disastrous results (thankfully it was a lab server). The strips just let go over time.
    – Jim B
    Jan 3, 2012 at 21:25
  • Good to know. So, you heard it from @JimB: use the real bracket. (I've only used 3M Command adhesive strips to mount SSDs, which of course don't vibrate.)
    – Skyhawk
    Jan 20, 2012 at 17:08

Though I don't have experience with this configuration, I think the overall approach is valid.

It is true a laptop will power its drive down so power-wise it may be a wash. But keep in mind in this case dust is also an issue. A system designed to be fanless would hopefully have some kind of heat sink configuration to mitigate the lack of a fan to move air over components to cool them. And less moving parts means less likely to have things gum up with dust in them in general.

Another potential advantage with flash-based storage is that these types of locations generally also have poor power quality/availability. In theory having solid state storage will minimize the impact of power glitches vs. rotating cylinder storage. Though I would still invest in a UPS or at least strong surge protection of some kind if at all possible. Mostly because it will probably repay its cost in not losing systems regularly when power goes wonky and then having to ship them all over the place.

Space-wise, 32GB is probably sufficient.

You might also consider looking into caching the login credentials local to the specific office that needs them (vs. using an RODC). This can be accomplished on local machines and also on Windows 2008 R2 server, depending on configuration. This decreases security for one system/user/office, but then the only locally stored authentication information will be restricted to local users instead of potentially encompassing your entire organization. Probably not a major issue, but thought it was worth throwing out there.

  • I wonder if something like the Shuttle XS35V2 might be good for this? I use them and love them in HTPC situations.
    – Kyle
    Jan 9, 2012 at 17:28

Laptop HDs can be spun down when not in use wheras a lot of flash doesn't ever power down so heat/powerwise any advantage can entirely depend on the specific products and the usage pattern of the machine. Likewise with reliability. It doesn't sound like you are going to see the major benefit of shock resistance.

Also, even with a fanless system you will still build up heat and particularly if this is on all the time with high ambient temp you will need heatsinks/a heat dissipating chassis that can maintain the running system within operating limits (remember at high ambient temps you haven't got to just worry about the traditionally hot components but the rest will also have a shorter reliable lifetime)

Really you need to put something together and test it in as near to the expected conditions as possible.

  • 1
    Yeah - we would certainly pilot this, but if someone answers the question saying "we tried this already in Nairobi and it melted through the floor" then I won't bother :-)
    – dunxd
    Jan 3, 2012 at 17:34
  • I've worked on fanless systems that overheated in cars in the UK so cooling will definately be a consideration to be dealt with to some extent
    – JamesRyan
    Jan 4, 2012 at 9:47
  • Interesting. Do you know what the ambient temperatures where when the overheating occurred? Ouagadougou is probably our hottest location - max temperature in shade recorded is 48°C!
    – dunxd
    Jan 5, 2012 at 13:09
  • 1
    Only around 40C. This might be helpful download.intel.com/embedded/processors/thermalguide/323439.pdf
    – JamesRyan
    Jan 5, 2012 at 14:50

Your ideas are sound, but you should consider those RODCs to be disposable. If one does experience a hardware failure (especially Storage), you will need to forcefully remove it from AD, but this is not terribly complicated.

Check your current DCs for their memory usage, but 800 users isn't going to be terribly huge. Off the top of my head, you'd probably be fine with 2GB of RAM, but 4 is obviously better.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .