Upsetting the boat a bit

[zzing123]

Hi Tim,

While I appreciate the stringent rules you put NAS's through to try to keep a homogenous set up utilizing the same drives and network set up and the like, I was wondering whether you could begin testing a couple of other factors too.

- Multi-usage, even in home. When, say you are watching a movie from a NAS, while recording to a NAS from a PVR, doing a backup and working out of a files directory shared off a NAS all at the same time is not atypical. Certainly not in my house. In fact, 2-3 simultaneous usages it quite average. I think it's important to test and see how badly NAS's get bogged down in such cases.

- With the advent of 10gig NAS's, we should see how NAS's behave when more than one user is accessing data. I have a 'smart switch' (Linksys SGE2000) but want to know when it's worth upgrading or not. Conversely with smaller smart switches like Netgear GS108T's being ridiculously cheap too, shouldn't you be proposing that any self-respecting NAS owner has a LACP-capable switch?

- Naming, shaming, laziness and shoddy support. The prime example was QNAP having a kernel problem that is in dire need of proper exposure: http://forum.qnap.com/viewtopic.php?f=189&t=51741. There has been very little naming and shaming going on, which has led to QNAP getting away with shoddy support for far too long. If a NAS maker needs to be rapped over their fingers, do so! My QNAP TS-439 had suffered from 3 RAID failures (because mdadm wanted to - the same drives (Hitachi Deskstars) are still absolutely fine in my new solution), the kernel problem above and then the DOM failing the day after warranty ran out was the straw that broke the camels back for me with QNAP.

- Capacity-full and near-capacity-full problems. A lot of storage, especially these cheap NAS's begin to have serious problems with performance degradation at or above 80-85% capacity. Beyond the spindle slow down for sectors on the outside of the platter and fragmentation, there's no really good reason for this, but I'd expect NAS makers to compensate for this (if at least kick in defragmentation). Similarly, we should see what happens when it is absolutely 100% full, and how the NAS begins to fail. Even ZFS for example gets its knickers in a twist in this scenario. Most of this falls down to mdadm needing some capacity to manage the RAID, but it can be tuned, which leads me to...

- Proper tuning, experiment and using the right technologies for the job. I expect NAS makers to experiment and make use of new technologies for me, so I'm not the one risking my data to see what technologies would be best in my case. This means three things: 1) Tuning technologies properly. If anyone should be an expert on tuning the performance of Samba, iSCSI and mdadm, it should be the NAS makers. They all too frequently don't switch tunables to what is actually optimal for their own NAS's! 2) Experimenting with different file systems other than ext4 (In some cases Reiser and XFS are better, though agreeably ext4 is a very good 'default' until Btrfs gets its act together) and 3) Differentiation through advanced technologies like Flashcache and bcache (which is now nearing kernel inclusion:http://www.phoronix.com/scan.php?page=news_item&px=MTEwMDg),. The point is that Drobo has taken 'data aware tiering' to heart, and while Thecus did experiment with FUSE-ZFS, these efforts should be praised. Drobo has it's own problems (performance) and FUSE was a bad idea. SSD's and RAM in a NAS should matter more than they currently do. I think it's time for NAS makers to do so.

- Criticism to the general cost of NAS's, which is basically set at $100 per hard drive bay regardless of how many bays NAS's have is getting a bit tired. In reality, just how much more expensive is it to produce a TS-859 versus a TS-459? Don't NAS makers realise that the more bays a person has managed by a particular brand the more tied they are to their solutions going forward?

- Stacking. Why is only Synology allowing some limited form of expansion? It's ridiculous to assume that the boundaries of a NAS is limited to the metal shroud. Switches have been stackable since the 90's, it should be the case with NAS's too.

Anyway, sorry for the rant, but I'm sick and tired of seeing NAS's that cost more than a cheap PC (especially when you go to 8 bays and above) being supported so badly and simply not doing anything of any value to anyone but themselves. Can we shake the boat up a little and give them a bit of a kick up the behind please?

 

[thiggins]

You raise some good points, some of which I may investigate. I'm afraid increasing the test load for every product probably isn't going to happen anytime soon. Just don't have the manpower to do it.

Sounds like you are way more knowledgeable than I some of the topics such as the tuning and capacity issues. Maybe you'd like to write an article or two?

NAS pricing is set at what the market bears. As long as folks vote with their dollars for the more expensive NASes, that's how they will be priced. WD, Buffalo and Iomega do provide lower cost alternatives, but they have their tradeoffs. For higher-end NASes, remember that the alternatives that QNAP/Synology/Thecus, et al are looking to displace (HP, Dell, NetApp) are much more expensive.

 

[tipstir]

He does have some valid points about the NAS industry. Still the other choices are running file server 24/7 an add about 5 to 6 HDD in there. But the FSB has always been a bottleneck with 100MHz M/T. Currently now more UMI 5 GT/s along with 10/100/1000/10000 on some. Don't suffer from all they extra NAS issues. NAS mostly for backup and storage. The media part might be too much for these NAS system.

Thanks for sharing about the Netgear Smart Switch being so long in price, I would be more interested in the PROSAFE 16-PORT GIGABIT SMART SWITCH
GS716T-200 or PROSAFE® 24-PORT GIGABIT SMART SWITCH
GS724T-300.

 

[sfx2000]

Great post and great thread - one thing missed - power consumption

I think most of the userbase here is either home networking, or small office/home office, and having a box on 24/7...

 

[zzing123]

Thanks for you kind words to what otherwise was a bit of a rant! I'll consider your offer for an article, but my problem is that I have very limited resources to be able to experiment with to investigate properly which tunings and techs would be best to use - but I'll see what I can do.

With the enterprise, what people are really paying for is the ability for people to take them out to lunch and also agree to some ridiculous terms in warranties and the like. 'Enterprise' grade software is also another con for another time. As a developer myself I cannot see how 'enterprise' software requires customers to take out a consulting contract costing twice as much as the license just to install the damned thing. 'Enterprise' imo means product not fit for purpose that requires considerably more evaluation and testing to make consumer ready - aka 'lack of quality'.

Power consumption using these Atoms is already pretty good - even the QNAP TS-439 is 18W at idle, and at load you just have the HDD motors on top of that in reality, but you are sacrificing performance quite a lot due to the power envelope.

What annoys me is that for $775 you can get this:

Intel® Server Board S1200KP: $147.99 (Amazon)
Intel Xeon E3-1230V2: $235.99 (Newegg)
Intel® RAID Controller RS25GB008: $159.99 (Newegg)
Kingston KVR1333D3S8E9SK2/4G RAM: $34.99 (Newegg)
Fractal Design Array (PSU inc): $197.99 (Newegg)

Which will provide you with a hardware RAID platform good for 8 drives (though the case is only for 6 drives). In fact if you were using ZFS you can also make use of the 6 on board SATA connectors too, meaning this set up could power 14 drives in total (4 being SATA3). Being a Xeon E3, it's also got ECC RAM and can actually do something like run VMWare ESXi to be a useful server (you may need a little more RAM). All in a box that's pretty much the same size as a 6-bay NAS give or take.

OK, depending on which OS you use (FreeNAS, Windows, Solaris, Openfiler etc), you may not have as glitzy a UI, but the reality is this is storage. When has storage ever been glitzy?

I'm not including drives because you don't get drives in a TS-659. Obviously you don't need a Xeon in your NAS - you can make do with a Core i3 just fine too, which would bring the price down considerably.

The point is though, that comparing this $775 setup to a QNAP TS-659 priced at $950 it begins to beggar belief. I've just specced a full ECC and Xeon-based server with hardware SAS RAID that can be a useful server for a fraction of the cost of a non-ECC, Atom and Software RAID QNAP that can only be a NAS. OK I'd expect power consumption to be slightly (but not too much) up on the QNAP, but I doubt that you'd spend more than the $175 difference on power throughout the lifetime of the device, then by running VMWare, if you have other servers, this box could replace other boxes too, making the power savings better. It's all about perspective and usage scenario.

The main point though is that on price, the NAS makers are seriously taking us for a ride.

 

[tipstir]

I am just going to stick it out with true blue NOS just don't see the point spending so much for NAS where it's still consider slower than using NOS. Technology on the FSB is now pushing G/T gigabit instead of M/T megabit. Less bottle-necking!

Power used is another story but what are you going to do? You need data and file access to store then backup it. You also want quicker access to your data also.

To transfer 5 GB of data files over the network shouldn't take hours it should just take a few minutes less than 2 depends on the traffic, but still I just don't like waiting. Moving 650GB of data does take a good time though. But not a day or a couple of days like it would with WD 3 TB NAS ouch. That's bad to sell a product like that to the general public. Charge over $249 for it on their site.