Commercial NAS Operating Systems - A Comprehensive Overview of Core Features
by Ganesh T S on November 14, 2016 8:30 AM ESTSetting the Stage...
As we draw the first part of our comprehensive coverage of commercial NAS operating systems to a close, it is important to touch upon one additional core aspect. The setting up of multiple users, user groups, joining domains etc., LDAP or NIS authentication etc. are very important for business NAS units, but, not so much for NAS units targeting the home market.
Since the COTS NAS operating systems are all based on Linux, adopting the standard user / group strategy is not very difficult for the vendors. ZyXEL, as expected due to their targeting of home users, has only very basic user and group settings with quota support. Western Digital goes a bit further by allowing for multiple users to be created at the same time. Asustor, QNAP, and Synology have that feature and much more too. For example, Synology even supports 2-factor authentication for certain groups. Netgear does support some advanced features like Active Directory, but there are lots of things that Netgear could learn from the aforementioned vendors.
Today's piece dealt with the core aspects of NAS operating systems - storage and how it is configured, the user interface, setting up of the desired services and shared folders, and the configuration of the network links. Even though the coverage has been very subjective, there are some clear areas for each vendor to improve.
Asustor, QNAP and Synology have the setup process nailed down to a decent extent. However, Netgear needs to make its ReadyCLOUD process more robust. Alternatively, the RAIDar program should be fixed to avoid Java requirements. Western Digital's approach is almost perfect, given that they mostly sell systems with disks pre-installed. However, it would be good if a volume is created by default when the My Cloud OS is installed. ZyXEL's approach is passable, but the slow web UI leads to an unsatisfactory UX.
On the storage and services side, Synology and QNAP turn out to very feature-rich, followed closely by Asustor. Netgear still has some catching up to do as certain aspects like advanced SMB options still require an external package to be installed.
In terms of networking features, QNAP is very much on top. While all vendors have some sort of teaming implementation, QNAP has gone beyond that and started to implement various network modes that can really take advantage of the multiple LAN ports.
Next week, we will have a follow-up article that deals with value-add features. These include media services, surveillance (DVR for IP cameras) solutions, and the public cloud (integration with Dropbox, Google Drive etc.). We will also discuss support for virtualization - in terms of being a datastore, as well as the NAS being a host for guest VMs. A look at some of the third-party applications and the usage models that they enable will round up our comprehensive coverage of NAS operating systems.
Facebook
Twitter
RSS
103 Comments
View All Comments
cdillon - Monday, November 14, 2016 - link
I meant to say "data and metadata check for ZFS" above, because it does both by default. You can turn off data checksums in ZFS just as easily as you can change the checksum algorithm, but can never turn off metadata checksums. ZFS will also never store less than 2 copies of all metadata (optionally more), giving you built-in redundancy in that respect, even in single-disk setups. This is one of several reasons that ZFS doesn't have and doesn't need a "fsck" utility, because the filesystem integrity is baked in to the design.tuxRoller - Thursday, November 17, 2016 - link
Btrfs has actually supported sha256...TWICE. The first time was back during early btrfs development and then they removed it because of how slow it was. It was submitted in 2014 by a oracle dev but not merged because they want a general solution, and, potentially, much more flexibility (using the crypto API instead of having to maintain their implementations;per file hash algorithm (so, some files could use a stronger hash than others); different hash functions for data and metadata (metadata is limited to 256 while data doesn't really have a limit))It's been percolating its way towards the kernel but it's not seen as a huge priority because: 1) there've been very few incidences of corrupt blocks passing ("crc32's error rate works out to one false positive per dozen megabytes *of random errors*--- that's a lot of errors, even talking into consideration CERN's data), 2) they already use sha256 for dedup (both in and out of band, though groundwork has been laid to make that pluggable as well).
linux-btrfs@vger.kernel.org/msg39109.html" target="_blank" rel="nofollow">https://www.mail-archive.com/linux-btrfs@vger.kern...
linux-btrfs@vger.kernel.org/msg39162.html" target="_blank" rel="nofollow">https://www.mail-archive.com/linux-btrfs@vger.kern...
linux-btrfs@vger.kernel.org/msg39205.html" target="_blank" rel="nofollow">https://www.mail-archive.com/linux-btrfs@vger.kern...
linux-btrfs@vger.kernel.org/msg50364.html" target="_blank" rel="nofollow">https://www.mail-archive.com/linux-btrfs@vger.kern...
ghostbit - Monday, November 14, 2016 - link
Nah. I am very happy with the Web interface and iOS apps Synology offers along with a sleek and small low-power chassis.But I guess I should also be ashamed I am not running a desktop with pfSense as a router as well?
dave_the_nerd - Monday, November 14, 2016 - link
Aside from the pfsense / FreeNAS issue - if you're using old desktop hardware as a server, yes, you should be ashamed. :-PIf you are using a prebuilt distro like FreeNAS instead of rolling your own distro, you should also be ashamed. Or something.
MrCrispy - Monday, November 14, 2016 - link
ZFS is expensive and hard. Can't mix and match drives, RAM hog, to expand a vdev you need to upgrade all disks, needs ECC ram, keeps all disks spinning, will slow down as you near capacity.The ZFS hype needs to stop. Its not at all suitable for a home user NAS, its meant for data centers.
bsd228 - Tuesday, November 15, 2016 - link
some of us don't see a difference, Crispy. I value my data as much as my datacenters' data.doggface - Tuesday, November 15, 2016 - link
Pfft. RAID is dead. Any decent array these days, is likely to get data corruption from hw raid.ZFS is rock solid. Also, if for some reason your OS/COTS solution dies your data is at significant risk of death. ZFS datasets are kept locally and you can rebuild the dataset regardless of the hardware.
eldakka - Wednesday, November 16, 2016 - link
RAID is dead? Huh?Sure, you can use ZFS without RAID, but that doesn't give you any hardware failure data protection.
I've rarely seen someone not use ZFS in some sort of RAID arrangement:
ZFS RAIDZ = RAID5
ZFS RAIDZ2 = RAID6
Software RAID (ZFS RAIDZ, RAIDZ2 and others) is still RAID.
doggface - Wednesday, November 16, 2016 - link
Ok. You got me, yes, it is uses an array of disks. Let me be more specific, RAID5 is a risk not worth taking. The reason they call it RAIDZ and RAIDZ2 is because they are significantly different to normal raid 5/6. Any RAID5 array that is over 12tb total size (not pool size) is at a very high risk of corruption as soon as a disk dies, its not conjecture - it is math. ZFS mitigates the risks inherent in a system designed decades ago, and is a far superior option for massive and small deployments.eldakka - Friday, November 25, 2016 - link
OK, then let me be more clear.RAID = Redundant Array of Inexpensive Disks.
ANY technology that uses multiple disks and incorporates some level of redundancy such that the ARRAY spread across MULTIPLE DISKS that can withstand the failure of 1 or more of those disks (you know, REDUNDANCY) IS a RAID array. Whether it's in software or hardware, whether they are fixed size stripes or variable sized (ZFS uses variable stripe sizes so it can optimize the size of the striping to each individual file) stripes, whether it uses stripes or separate ECC disks or complete mirrors, it is ALL a subset of the RAID paradigm.
ZFS RAIDz, 2z etc are RAID arrays. Their implementation of adaptive striping sizes per file is a significant enhancement over standard RAID5 and 6 and so on, but it is just that, an enhancement of, or more sophisticated implementation of, RAID.