Benefits of NAS

[roaf85]

I was looking into getting a NAS enclosure. I was wondering what is the advantages of them over a regular server. I would think for cheap the performance of a dedicated server would be better than a cheap NAS solution, but what about the power usage? I would think NAS would have a big advantage in power usage.

This is home computing by the way, but just trying to get some opinions on TCO.

[swatbat]

A nas device is easy to setup and yes prob uses less power.

Server based on a normal computer lets you do more and can be more expanable. Generaly speaking it is harder to setup although if you where using something like windows xp it would be pretty easy. Hell freenas is easy to setup as well. PC's performance can be a lot better or about the same depending on how much you put into it.

How many drives you looking to run?

[Cyrilix]

I was looking into getting a NAS enclosure. I was wondering what is the advantages of them over a regular server. I would think for cheap the performance of a dedicated server would be better than a cheap NAS solution, but what about the power usage? I would think NAS would have a big advantage in power usage.

This is home computing by the way, but just trying to get some opinions on TCO.

The advantage is simplicity, integration, and size. For me, integration and size were the big ones. Simplicity of a web interface was just a bonus, sort of like configuring a simple router.

[Madwand]

I was looking into getting a NAS enclosure. I was wondering what is the advantages of them over a regular server. I would think for cheap the performance of a dedicated server would be better than a cheap NAS solution, but what about the power usage? I would think NAS would have a big advantage in power usage.

This is home computing by the way, but just trying to get some opinions on TCO.

There is nothing about "NAS" per se which guarantees low power consumption, and there is nothing about a "regular server" which guarantees high power consumption. These are essentially design choices / trade-offs. A "NAS" could be a massive powerhouse, and a "regular server" could be a very efficient low-power and small form factor design. Look around, and you'll see both of these in actuality.

The big difference, IMO, in your usage, is that one implies an off-the-shelf consumer device, and the other a DIY server build of some sort. The latter is more open-ended, and can in particular be built to incorporate significant power savings through automated sleep / wake up if not at all times.

The big difference between these two approaches is that in one case, most of the decisions are made for you and fairly fixed and limited, and in the other case, all the decisions are your own. The advantages or disadvantages of these two approaches are pretty much in the eyes of the beholders -- to be evaluated based on their preferences and limits of knowledge / time / effort / skill.

I happen to be a fan of the DIY approach, because I find the trade-offs imposed by consumer NAS boxes to be too limiting and costly for the performance and advantages gained, but think they're reasonable solutions for those who don't wish to spend the time/energy and prefer pre-built, or don't place much importance on performance.

[Cyrilix]

There is nothing about "NAS" per se which guarantees low power consumption, and there is nothing about a "regular server" which guarantees high power consumption. These are essentially design choices / trade-offs. A "NAS" could be a massive powerhouse, and a "regular server" could be a very efficient low-power and small form factor design. Look around, and you'll see both of these in actuality.

The big difference, IMO, in your usage, is that one implies an off-the-shelf consumer device, and the other a DIY server build of some sort. The latter is more open-ended, and can in particular be built to incorporate significant power savings through automated sleep / wake up if not at all times.

The big difference between these two approaches is that in one case, most of the decisions are made for you and fairly fixed and limited, and in the other case, all the decisions are your own. The advantages or disadvantages of these two approaches are pretty much in the eyes of the beholders -- to be evaluated based on their preferences and limits of knowledge / time / effort / skill.

I happen to be a fan of the DIY approach, because I find the trade-offs imposed by consumer NAS boxes to be too limiting and costly for the performance and advantages gained, but think they're reasonable solutions for those who don't wish to spend the time/energy and prefer pre-built, or don't place much importance on performance.

Hmm, well, performance certainly isn't terrible. I get 30 MB/s for my standard 100 MB files. I don't imagine most consumer devices would do all that much better. The network seems to be, somehow, the NIC, no matter what kind of NAS you use, DIY or prebuilt.

Also, size is a big one! There's no way you'll get a box as small as a small toaster if you DIY it. Even Mini-ATX are huge compared to the size of these boxes (assuming 4-5 hard drives). If you want a case with more hard drives, then it's more or less the same, however.

[Madwand]

Hmm, well, performance certainly isn't terrible. I get 30 MB/s for my standard 100 MB files. I don't imagine most consumer devices would do all that much better.

30 MB/s is actually not bad, but I'd want to double-check the details of the hardware and the test. Small file sizes can give misleading results due to caching effects. SmallNetBuilder for example has measured many consumer NAS boxes which don't come close to 30 MB/s according to their tests when you factor out the misleading small test file results.

FWIW, I've exceeded 100 MB/s actual very large file transfer with consumer gear using Windows file transfer, and don't find it hard to make a DIY build which exceeds the performance of any NAS device as measured to date by SNB for their NAS charts, using the same test method.

[Cyrilix]

30 MB/s is actually not bad, but I'd want to double-check the details of the hardware and the test. Small file sizes can give misleading results due to caching effects. SmallNetBuilder for example has measured many consumer NAS boxes which don't come close to 30 MB/s according to their tests when you factor out the misleading small test file results.

FWIW, I've exceeded 100 MB/s actual very large file transfer with consumer gear using Windows file transfer, and don't find it hard to make a DIY build which exceeds the performance of any NAS device as measured to date by SNB for their NAS charts, using the same test method.

That's odd. I find small file size transfers to be the worst-performing, whether it's copying from an internal hard drive to another internal hard drive, or otherwise. If I were to quote you my speeds with transferring many many small files such as small pictures, they wouldn't be nearly as good as the one I mentioned above.

[Madwand]

That's odd. I find small file size transfers to be the worst-performing, whether it's copying from an internal hard drive to another internal hard drive, or otherwise.

This is correct -- small files actually transfer at slower speeds than large files, because the proportion of overhead to data is higher. However, small files can easily fit in the RAM cache, so when you only use small files for testing, you can get misleading results because of speedups due to pulling files from cache or only writing files to cache and not the actual hard drives. This can be seen in SNB's tests among others, which lead me to a general mistrust of tests involving small files. Rather than trying to ensure that caching effects have not polluted the results despite using small files, I prefer to use very large files for which caching effects would be relatively small.

Here's a similar recommendation from the Bonnie benchmark for example:

http://www.textuality.com/bonnie/advice.html

It is important to use a file size that is several times the size of the available memory (RAM) - otherwise, the operating system will cache large parts of the file, and Bonnie will end up doing very little I/O. At least four times the size of the available memory is desirable