Windows PC versus Windows NAS

Topic

New Reply

I currently have a regular Windows 7 PC which I use as a file server. I’ve been told that for security reasons I need to shut it down and purchase a NAS. I thought that the NAS might run some kind of more secure operating system which might make this true. However, today I found out that I can only purchase a NAS that runs Windows. How is this different from just running Windows on a PC? The only search that I was able to do on the web indicated that a NAS might be cheaper, but this does not appear to be the case, especially given that I already own the PC. I might have to do it whether or not it makes sense, but I’d just like to know if there is an rationale for the switch in hardware. Thanks.:confused:

Reply | Quote

Replies

I can’t see any reason for abandoning Windows as your NAS OS, assuming it’s not internet facing. Why do you think you need to change?
If you really want to change to a more secure unit replace Windows with FreeNAS on your existing PC, but this will require you reformat your disk(s) – lose data if you can’t backup.

cheers, Paul

Reply | Quote

I am using Windows 7 (and before that Vista, WinME and even back to Win98 and Win95) as a file server, which does the job of a NAS. I have it set up to be simply the OS, with advanced file sharing activated, so that I have control over which user can see which folder (and its files). This is so simple to set up and manage.

With respect to firewalls, the built in Windows version works for me – but since I use Sophos Endpoint Security, it comes with another firewall, which is also in place. Access to the unit is over the local LAN, but if I really get to need the files when I am away, then I use DynDNS to find my ADSL router. Then using NAT, I get to access the machine using Remote Desktop and from there the files I need.

The advantage of Windows as File Server is that you can install other applications to keep track of things (such as network traffic – both local and internet) and productivity tools so you can edit or change documents / spreadsheets etc directly on the server (using RDP and an iPad for example!).

I have looked hard at the NAS option, and even bought a cheap version, but never used it. The difficulty I have found is that the NAS usually runs a version of linux which stores files in a different file format to Windows. If you need those files, or the NAS dies, you can’t physically extract the hard drive and access the info. You can do that with Windows (as long as its not encrypted).

My 2cents: Stick with Windows, and make sure the OS is properly secured and patched with updates. Keep a good firewall in place, and you’re good to go!

Reply | Quote

Windows can access a Linux disk, it usually uses a format called Ext2.
Running the NAS on an old PC has the advantage that hardware failure is not the end of your data.

cheers, Paul

Reply | Quote

I’ve found an alternative that works well for me that combines both Linux and a Widows-friendly NTFS. I’m using OpenMediaVault with the OS loaded on a 60G SSD, and a 2TB WD Green label NAS drive formatted with NTFS. OpenMediavault reads and writes the NTFS just fine for my purposes (I use the bootable Linux version of Acronis 2013 to backup/recover over the network) Prior to using the SSD, I was running a 40G SATA drive. When I changed out the drives and reloaded OMV, all I had to do was re-attach my other drive and I was good to go. You might test it out and see if it works for you. IMO, OMV is MUCH easier to use than FreeNAS, which I looked at too, and doesn’t require near the resources that FreeNAS recommends.

I forgot to mention that I also have a shared folder on OMV that I use as a mapped drive on my Windows systems to freely copy files to and from without any problems after over a year of use now.:D

P.S. I’m also running a low-powered home server with Windows Home Server 2011 OS on a mini-ITX MB with Intel ATOM 1.6 GHz CPU and 2G RAM that is doing multiple daily backups for all systems connected to the network that might serve your purpose as well. WHS2011 was still available on Newegg recently for about $50 USD.

Best regards,
Phil

Reply | Quote

OMV doesn’t seem to support RAID Z. RAID Z is a major benefit on home systems IMO.

cheers, Paul

Reply | Quote

OMV doesn’t seem to support RAID Z. RAID Z is a major benefit on home systems IMO.

cheers, Paul

Ok I tried to look up RAID Z and it was a bit complicated to my fuzzy brain. In a few words what would the benefits be?

🍻

Just because you don't know where you are going doesn't mean any road will get you there.

Reply | Quote

OMV doesn’t seem to support RAID Z. RAID Z is a major benefit on home systems IMO.

It would be of far more benefit if the implementation weren’t brain-dead: to obtain the write integrity that you mention it requires that every write visit every disk in the RAID set (vs. perform read-modify-write actions to only two disks in the set as RAID-5 does) and, worse, that every READ visit all but one disk in the RAID set (vs. just one disk for RAID-5), thus dramatically reducing random IOPS for the array. Maintaining a transaction log and/or lazily propagating consistency information upward to parent nodes (which my vague recollection is that they actually do in other areas) would have accomplished the same goal far more efficiently.

Reply | Quote

It ensures write integrity so the parity data and the actual data remain valid in the case of write failure – power etc. RAID 5 doesn’t do that so you need a controller with a cache to store uncommitted writes. If you only use RAID 1 (a waste of disk space IMO) then it’s not an issue.

cheers, Paul

Reply | Quote

It ensures write integrity so the parity data and the actual data remain valid in the case of write failure – power etc. RAID 5 doesn’t do that so you need a controller with a cache to store uncommitted writes. If you only use RAID 1 (a waste of disk space IMO) then it’s not an issue.

cheers, Paul

If I am understanding correctly, if the parity is being written in Raid5 and does not complete due to a power failure parity and data could be lost?

🍻

Just because you don't know where you are going doesn't mean any road will get you there.

Reply | Quote

This was touched upon by pheberer when he mention using an Atom 1.6GHz processor. My point is that almost any old PC can be configured as a server. It doesn’t have to be powerful or have a lot of RAM memory. I set up a friend’s old computer which has an AMD 1.4GHz single-core cpu to use as a home server. It runs Windows XP. His household has several newer computers all of which run Windows 7 or 8.1 and the system is trouble free. Most of the time the old rig ticks along at around 800MHz processor speed and costs only pennies to run. It’s set up using the basic Windows Home Networking with all users having access to most files. The only exception is some protected financial stuff which “Dad” requires a password to open that folder. He doesn’t want his teenage children’s friends viewing that stuff! There are two 2TB hard drives in the case plus a small SSD which runs XP. It’s set up so that the second hard drive does a daily backup of any changes/additions/deletions to the data on the first hard drive. RAID is not required. The family doesn’t need remote access away from home so Windows firewall and AVAST anti-virus provides adequate security. This is a case of getting lots of extra mileage out of an old machine. Of course, his daughter told him she desperately needed an iPad which they could now afford because he didn’t spend money to buy a home server ….. 😀

Reply | Quote

I desperately need lots of things. 🙂

cheers, Paul

Reply | Quote

Do you have figures for relative performance?

cheers, Paul

Reply | Quote

Do you have figures for relative performance?

They are as I stated. For random Read operations smaller than the stripe chunk size (which was typically 128 – 256 KB a decade ago and is likely more now) a RAID-5 array of N disks can provide N-1 times the IOPS that a RAID-Z array can (or N times if RAID-Z is structured like RAID-3 witl all the parity on one disk – it’s been a while since I visited this discussion). For similar random Write operations a RAID-5 array of more than 3 disks (in which each random write takes less than the equivalent of 4 random accesses: two pairs of seek-wait-half-a-rotation-to-read-then-wait-a-full-rotation-to-update operations) may also out-perform RAID-Z (where each random Write requires a full random access seek-wait-half-a-rotation-to-update for every disk in the array), though for writes the latter’s drawbacks may be mitigated by ZFS’s no-update-in-place behavior which may be able to aggregate them. I well recall the outrage with which these observations were greeted by the ZFS faithful on Sun’s blog until both the principal ZFS architect and one of the ZFS engineers active on the blog confirmed them. You can still find one of the latter’s blogs on the subject (and my commentary on it) at https://blogs.oracle.com/roch/entry/when_to_and_not_to

Reply | Quote

Correct. RAID 5 does not guarantee RAID consistency until the data and parity write is complete. A failure between those writes can result in data loss.

cheers, Paul

Reply | Quote

Correct. RAID 5 does not guarantee RAID consistency until the data and parity write is complete. A failure between those writes can result in data loss.

Actually, that depends on the implementation. For example, a RAID-5 implementation that refuses to support write activity again after a power outage until it has reconciled all parity information with its related stripe data will not lose data in such cases (regardless of which was written first the write operation cannot properly have been considered to have completed until both have been written, hence it doesn’t matter which you choose as the valid one). This can be a viable mechanism in UPS-protected RAID-5 environments since power loss is such a rare event there; a status flag on the disk that is set on the first write to the array after start-up and cleared on a clean shut-down to differentiate between the latter and a power failure on restart. This can also protect against ‘bit-rot’ where data or parity becomes detectably invalid, since the vast, vast majority of bit-rot cases do not leave a sector looking valid. And once the array is already operating in degraded mode (missing a disk) the issue doesn’t arise. There are more obscure cases in which a multi-sector write to one of the disks is only partially completed on a power failure and in such cases RAID-Z does enjoy an advantage in robustness over RAID-5 – but RAID-Z COULD have been implemented to maintain this advantage without its drawbacks had it instead employed logging or parent-resident metadata techniques to guard its integrity (for that matter a modified RAID-5 implementation could as well and in fact often does when some NVRAM is available to use as a short-term log until each write has fully completed though I don’t know of one that uses a normal hard-disk-based transaction log for that purpose).

Reply | Quote