Articles - Question of the week: what constitutes a perfect NAS device?

[HEXUS]

Network-attached storage devices are becoming increasingly popular and all the more feature packed. But what really makes the perfect NAS device?

Read more.

[SiM]

Only 3 things needed gigabit, linux and reliablity

[nichomach]

Good lord - how long's a piece of string?

First and foremost, resilience and stability. Doesn't matter if it goes like the proverbial off a manual excavation implement if that just means it loses all your data REALLY quickly.

Next, compatibility. There are still companies chucking out kit that requires some piece of dedicated software to connect to it. Wrong. Ideally, it should support access via as many clients as possible, be they NFS, SMB, whatever. It should also not require anything more complex than a web browser to configure.

Security; it should be possible to lock down files on the system by user account, and ideally that should be possible through authentication against AD, eDirectory, NTLM etc.

Expandability - and that doesn't just mean increasing capacity with another disk. It should include the capacity to be enhanced with automated replication/backup capabilities, for instance. I have a Buffalo LinkStation Pro on one site that runs an unattended backup to a DriveStation hooked up to it via USB. Simple, yes, lots of devices do it, yes, but it's still worth emphasising.

Performance. Of course, I like fast. Who doesn't?

Beyond that, it gets a bit complex; for a NAS in a corporate environment, for instance, I expect a lot more in the way of performance, redundancy and management tools than I do from one in the home/SOHO environment. In the latter case, I might accord a higher priority to capabilities like media streaming.

[kingpotnoodle]

Fast, quiet, gigabit networking, good support for various RAID levels, ideally 4+ drives, low power consumption, web interface, does neat things like download management, audio streaming for you (for instance on my file server I run the open source Subsonic), sends alerts about hardware failure via email at minimum.

Now a cheap linux server like mine can do all the above, so it's gotta be low cost to complete with that, and to get one up on such a machine it'd need things like hot swappable drives, small size. very, very high uptime, quick boot times, easy and quick setup.

Like the above poster it would also be very nice to have security controls, different user names to access different segregated areas, and remote authentication as well, access across many protocols (SMB, FTP etc).

I also like the idea it automatically backing up, but for me off site would be the only really good way as it's RAIDed already to another drive attached via USB seems superfluous... so it should have remote sync capabilities with online storage providers or with another NAS in another location, and use your internet bandwidth sensibly, during off peak hours - if necessary to build up the backup over several days.

[aidanjt]

Here's my NAS wishlist:

• Modular WebUI Interface.
• A simple package manager to offer new services.
• Document the WebUI API and packaging system to allow end-users to add their own services not offically supported.
• One or two USB ports would be nice to add printing services, but not hugely important.
• A few eSATA ports for external storage expansion.
• Allow 3rd party firmwares (obviously make the user aware of garentee issues by using 3rd party firmwares), foster community development around your device, nothing makes an embedded device more popular for enthusiasts than good ole fashioned openness. One only needs to take a glance at the Linksys WRT-series routers for proof of this. And embedded hardware companies benifit from recieving bug reports and patches from firmware developers. Kernel developers always loves kernel patches from professional hardware manufacturers, so they'll maintain the kernel for you so you don't need to bug wrangle with internal patchsets breaking with new kernel releases.
• More CPU and memory horsepower to allow multiple services and reasonable responsiveness, maybe use a CPU supported by the Linux cpufreq sub-system. Even silly cheap x86 CPUs can achieve this and DDR2 chips are silly cheap.
• More drive bays, 8 would be nice
• More flexible disk management, i.e. use mdadm to swap in larger disks, reassemble, and grow the array as the disks are been replaced. Pretty straight forward using the raid456 kernel module.
• More competitive pricing, right now it's cheaper to build a new NAS with PC hardware to attach 6 disks than to buy an embedded 5-bay NAS with less disks space, memory, and horse power, a bit silly really. Sure you don't get the backplanes, but they're not hugely expensive either.
• Maybe UPnP AV on-the-fly transcoding to save the hassle of manual transcoding media for dealing with multiple target clients, maybe do a deal with AMD for a condensed variant of 'AVIVO converter' circuits to mitigate the CPU usage. Although this may be costly to develop, so it wouldn't be a dealbreaker to be without this feature.
• Dropbear is a tiny SSH daemon and can make secure transmission of files over the internet really easy and let power-users dig around.
• Offical support for common network data services like OpenLDAP and Squid proxy server would be hot, most NAS devices already have MySQL support so it's clearly already an easy task.
• Dual-gigabit NIC bonding is a must.

Basically it would be great to have network-in-a-box functionality from a NAS, rather than running a fully-fledged server eating more power, even if it ends up using nearly the same amount of power, the obvious convinence and space factor from an embedded device would still win me over. I have tons more ideas about the 'perfect' NAS, as well as implimentation ideas. I'd only be too happy to discuss more of them if Tecuas is interested.

[pumpman]

something really easy to use but cheaper than a Drobo

[Mackintire]

I 'm going to plagurize the first post a bit, because he was dead on for most of the things he said.

1. First and foremost, resilience and stability. If it loses your data, Its a failure.
2. 
   Compatibility: There are still companies chucking out kit that requires some piece of dedicated software to connect to it. Wrong. Ideally, it should support access via as many clients as possible, be they NFS, SMB, whatever.
3. 
   Ease of use: It should also not require anything more complex than a web browser to configure.
4. 
   Security: it should be possible to lock down files on the system by user account, and ideally that should be possible through authentication against AD , eDirectory , NTLM etc.
5. 
   Expandability - and that doesn't just mean increasing capacity with another disk. It should include the capacity to be enhanced with automated replication/backup capabilities, for instance. I have a Buffalo LinkStation Pro on one site that runs an unattended backup to a DriveStation hooked up to it via USB. Simple, yes, lots of devices do it, yes, but it's still worth emphasising.
6. 
   Performance: My minimum standard is RAID 5 with a minimum thoughput twice the speed of a single modern Hard Drive. What this means is I can pound the storage device with two machines at the same time without noticing a significant slowdown. Most better quality hardware RAID 5 controllers can do this. We are talking sustained transfer rates higher than 200Mbps.
7. 
   Media Server: For a home user having a Media server is important.
8. 
   Allow open development: Foster community development around your device, if someone wants to put their own software on the unit, why not. You'll just sell more.
9. 
   Polish and Support: If you design the unit as I have suggested you will have one of the most powerful and electrically efficient units around. Keep supporting and polishing the user interface until it is perfect and the hardware is finally holding it back. Its what's made Apple what it is today.

Hardware Suggestions:

• Use an ATOM or the VIA Nano as the CPU (some crafty end users could turn this unit into a mini server increasing its value).
• 
  Integrate a 3ware RAID controller and lighten the CPU load. Your performance level would jdramatically increase.
• 
  Use a Intel Pro 1000 nic connected to a PCIe bus with user switchable jumbo frame support (gigabit teaming not needed but would be a nice feature to have)
• 
  4 USB ports would be good (placing resetable fuses on those ports like gigabyte does on their motherboards would be better)
• 
  Use standard size fans or even better use FDB (Fluid Dynamic bearing fans). Or sell them as an option.
• 
  For any led lighting on the outside have a adjustment for the brightness in the Web interface. Some people like to see a nice bright LED, some people find it annoying.
• 
  For the overall size and form factor 4-5 drives is idea. (Imagion using five 1.5TB drives in a RAID 5) 6TB of storage!!! Some people may want more drives but 90% of those out there will not.

Auto expand replace your drive one at a time allowing the unit to rebuild between each upgrade. When all the drives are replaced the NAS will ask you if you want to expand to RAID to the new size available. I think drobo can do this but am unsure if the newest Thecus devices can.

[Funkstar]

I think all the most importal points have already been covered (reliable, expandable storage, afordable, etc.)

Building a user community round your NAS is also a very good idea, us geeks are a loyal bunch if treated right and can add a huge amount of value to your existing products. Thecus have this almost right. Additional software can be added through modules, leaving Thecus free to concentrate on the key features.

One area that Thecus is quite badle beaten by it's competitors (Netgear/ReadyNAS in particular) is its comunity management. ReadyNAS has a really good forums where the developers hang out and help with all kinds of problems. Thecus have support forums here which only Yvon oversees, there are two user forums (one may be unoficial, that might be because the official one uses terrible forum software) and a Wiki. It would be realy good if they were all together (especially if they were all here).

Thecus also have to make sure they are on the ball with regards to source releases. I've seen a couple of threads on here of people pointing out that Thecus are in breach of the GPL as sources for their latest firmware are not available.

[digit]

For me to consider a NAS the price is the most important factor, perhaps then followed by performance. With hard drives as cheap as there are, to pay a ton for the nas device is a bit offputting and I'd rather just wack said hard drives directly into my pc. The only other feature i'd really ike is media streaming which would be inexpensive feature to add.

Perhaps releasing a basic no frills device with an open linux based firmware could be the way to go, then get users to develop and add additional features as they see fit. Linksys with its NSLUG (or similar named device) NAS became very popular with its community once flashed to different firmware for being a basic device which then users built upon.

[GaryRW]

Operating temp! I want one in my loft where my network switch is (and where a burglar is unlikely to find it and nick all my family photos...). But I hate to think what temp that is right now, and it gets pretty cold in winter...

[quatermass]

Most of the replies seem to be for personal use.

For a small business I'd need a TCP/IP 1Gb/s with RAID 1 at a decent speed (most NAS are slow due to cost cutting in providing a small cpu). I want also to be able to do a hot swap on the drives easily.

Oh and it must have SAMBA on board to auth with a Windows 2003 server.