I did ponder this, as i do have a spare PC laying around doing nothing. I basically only need it as a filestore, without RAID so im sure i could get a linux build to do this. Might have to look into it further, shame tho as i really want the Thecus! :)
My DIY boxes can do > 80 MB/s actual (8.4 GB) file transfer over consumer GbE without jumbo frames, so I agree that it's unlikely that networking alone is the issue holding back performance here. (As an illustration of networking + file access protocol speed; there are differences in the drive setups.) I add that when the files are fully cached, I can see net transfer performance in excess of 110 MB/s across the networkQuote:
Originally Posted by Steve
In addition, trunking wouldn't work as people here seem to imply here -- it would only improve aggregate bandwidth for > 1 simultaneous connections, not improve a single data access.
I can't post a ULR yet, so I'll quote a bit from www . alacritech.com / html / teaming.html
nVIDIA and others say essentially the same.
In addition, trunking is 802.3ad, not 802.3ab as stated in the review (assuming I didn't misunderstand the point of reference).Quote:
There are a couple of common misconceptions about link aggregation that cause people to have unrealistic expectations of resultant performance. The first is that link aggregation automatically multiplies your network throughput by the number of links in the team, regardless of connection type, environment, etc. This is absolutely untrue. Link aggregation increases your total throughput (when set up properly), but will not increase an individual connection's throughput at all. That is to say; if you can do a file copy at 50MB/sec on a single gigabit link, if you create a team, your speed for that file copy will still be exactly the same. If you do multiple simultaneous copies your total throughput will increase, but the copy time for any individual file will either remain the same, or possibly decrease, depending on resource contention. The reason for this is that all of the common link aggregation protocols require that a given conversation must be carried across a single link at a time.
----------------
Thanks for the review and the focus on performance. Perhaps reviews like this will spur the industry to continue to improve the performance of consumer NAS boxes.
That suff about trunking performace, and there being no improvement unless you make multiple concurent transfers sounds fishy to me. Rember that the TCP/IP protocol used to transfer the files is packet based. The file is split in to thousands of packets of about 1-9Kb each, and transfered sepearately, before getting re-assembled at the other end, the re-assembly part should not care if some of the packets traveled different routes, just that they all arrive. It is normal out on the public internet for routers to expect some packets to travel different routes from others
Imagine pouring rice from a bag into a jar. The rice is the file and the hole in the bag is the ethernet connection. You can get faster transfer by making a bigger hole (upgrading from 100Mbs to Gigabit), or by making a second hole in the bag. Both are effective at getting rice into the jar, and as long as all the rice ends up in the jar, you don't care how it got there.
If on the other had the file was transfered via a packet less protocol (they have not been used in computers since the early 70's) then nVidia would have a point. Suppose you add a second telephone line to your house, you can't use it to speak to someone faster, but you can speak to more than one person at once.
Seeing as the filetransfer is not stream based, this sounds much more like a limitation in nVidia's drivers (or hardware) or perhaps with windows.
It's fine to dispute the claims, but in this case, you should dispute them with performance measurements that back up your claims. I've followed this through, and that's what I've learned and seen in practice with trunking that I've set up.Quote:
Originally Posted by chrestomanci
If trunking was so easy to do, and give such great performance improvements, we'd all be doing it. All we're seeing now is a few products that put multiple NICs in consumer products, and people are getting excited about them, thinking that they'll do something that they won't. Trunking is great for multi-client servers, but are not generally useful for home users and single-user scenarios.
Post some explicit configuration detail and benchmarks if you want to dispute my claims.
I am not saying that it is possible to get better performance on nVidia hardware than they claim, I am saying that if what they say is true, it is because they have not tried hard enough to get it working, rather than there being some fundamental engeneering limitation.
If a file transfer is send via a single network interface, then allmost all the packets will arrive in sequence, making re-assembly very easy, but if they arrive via multiple interfaces then the re-assembly will be hard, as they will almost certainly arrive out of sequence. I suspect that Microsoft or nVidia have written the driver so as to refuse to allow more than one interface to be used for each file.
Link aggregation is done in L2, so it is not just a question of the particular TCP/IP implementation not "trying hard enough". The governing standard is IEEE 802.3. Clause 43.2.1 (f) says:
When the applicable standard says "all frames of a given conversation are passed to a single port", you can hardly blame nVIDIA or Microsoft for not "trying hard enough" in their implementations. Perhaps you can blame IEEE 802.3, but I think us doing so would be pointless and foolish.Quote:
Frame ordering must be maintained for certain sequences of frame exchanges between MAC Clients (known as conversations, see 1.4). The Distributor ensures that all frames of a given conversation are passed to a single port. For any given port, the Collector is required to pass frames to the MAC Client in the order that they are received from that port. The Collector is otherwise free to select frames received from the aggregated ports in any order. Since there are no means for frames to be mis-ordered on a single link, this guarantees that frame ordering is maintained for any conversation.
Linux's bonding implementation has an option that violates this rule -- balance-rr. On this, they say:
http : // linux-net.osdl.org /index.php/Bonding
So this is the problem -- the retransmission chews up bandwidth, and often defeats the purpose of the bonding. When I tried it, with 2 NICs, I got 1/2 the bandwidth of a single GbE link, and obviously concluded that it was better to use a single NIC.Quote:
This mode is the only mode that will permit a single TCP/IP connection to stripe traffic across multiple interfaces. It is therefore the only mode that will allow a single TCP/IP stream to utilize more than one interface's worth of throughput. This comes at a cost, however: the striping often results in peer systems receiving packets out of order, causing TCP/IP's congestion control system to kick in, often by retransmitting segments.
The previously "linked" Alacritech notes also entertain this point:
Again, theory's good, but practical performance figures are what count here. I invite you to try it and report better results and your methods, in any accessible implementation, including Linux, with any accessible hardware, that maintains minimal compatibility with standard ethernet, TCP/IP, standard file transfer protocols, etc. These are not arbitrary requirements; merely expressions of common expectations.Quote:
Round Robin Packet
Packets are distibuted iteratively across all the ports.
Advantages: Traffic is evenly divided amoung all the ports in the team, maximising throughput.
Disadvantages: Completely violates protocols in that a conversation isn't kept on a single port, and packets are not guaranteed to arrive in sequence.
Recommended for: Amusing yourself about how badly you can break your network. Never ever use this. It is a disaster.
I sent this email to Thecus support, will let you know if they get back to me:
Dear Thecus Support
I purchased your N5200 model after reading a positive review of it on Hexus.net. My requirements are as a home user, backing up my DVDs and storing HD content. I would also like to add that I'm a software engineer by profession, so have a good understanding of technical details and user interaction. I purchased the product for 805€ from nwkomp.de.
I would also like to make clear I will be posting this email on the relevant newsgroups and comments on reviews. I feel I owe it to other potential buyers. To be fair, I will also post any response from Thecus - so I hope it is a good response.
I received this item today (after a 2-week delay from the supplier), and I have to say I'm disappointed. I would appreciate a point by point response to the following:
1) Firstly, overall documentation is pretty bad - apart from basic grammatical and spelling errors, it's not actually very informative. As one example, out of many, the details regarding "user percentage" is very unclear.
2) Even the device itself, on the LCD, has spelling errors. It says 'formattin' instead of 'formatting', even though there are clearly several cells left on the display. Either someone noticed it and couldn't be bother to do anything about it, or simply didn't notice after, presumably, months of development. I am not sure which is worse. This single issue is not a great problem in itself, but I do think it indicates the quality of your products is less than great.
3) Here's a big qualm: Why does it take five hours to format 5 250Gb drives? Can this not be done in parallel? If it is done in parallel, that's even more shocking. This is supposedly in 'rebuild speed:high' mode - I shudder to think how long 'low' takes.
4) Here's the biggie that prompted this rant: Clicking 'remove' on the RAID page does exactly that. What it DOESN'T do is WARN you, or give you a second chance!!! That's it - BOOM, data gone. That sounds like a schoolboy error to me - one of the first rules of software interfaces is to protect the user. Fortunately, it IS what I wanted to do, but it could easily not be in the future, and I'd be VERY wary of even viewing that page, for fear of accidentally clicking the button, or pressing return with the button highlighted... Of course, its taking ANOTHER 5 hours to build the array - surely not... Thought I'd take the time to write this while it's chugging along.
5) The reason I chose to remove the RAID array was because there is NO WAY to change the percentages of Snaptshot and USB allocations. (By the way, to allow only specification by percentage is plain lazy. Why not absolute values? I think a lot of users would want a less that 1% amount dedicated to USB for instance, say enough for a 4gb card). Why can't you just have a special folder, which can vary in size like any other folder?
6) (This list is getting really quite long for a product I've not even finished setting up) The documentation clearly states that the red LEDs indicate a problem, and yet, from what I can gather from using the N5200, they are also red during formatting, except the top one, which is green. Explaining properly to the user what to expect during different states is again another basic error.
7) The web interface does not contain any context sensitive help! How easy would it be to put a little question mark icon, or something similar, that one could click to see what that particular item does?
8) The 'snapshot' item in the menu does nothing when I click on it (after I'd setup the RAID array the first time, and the system was reporting 'healthy')
9) Whomever wrote the content for the web interface does not appear to have had knowledge of the past tense in English. Again, not a great problem, but it's very irritating to see so many errors on a product I have spent so much on.
10) The 'status' column on the disk information page contains only the text 'OK?'. I click that to see nothing but 'N/A' for all the so-called SMART INFO.
11) Generally, the options and information all seems very limited. For a product of this expense I would expect to see far more information on the component disks, and generally spped and transfer amounts and rates.
12) The fan is really rather noisy. There is not need for that level of noise in such a small product. I know there will probably be some stock excuse, but to me in seems like poor engineering, considering the availabilty of quiet fans, and the fact that hard drives don't really need a great deal of cooling.
13) My main two reasons for purchasing this product, rather than building my own with a Linux distribution, is the small form factor of the case, and the plug and play aspect. In the first aspect, I am pleased - it's a reasonably compact unit, and is well presented. On the later though, I am really very disappointed, Sloppy documention, sloppy web interface, excruciatingly slow RAID setup, and just a general lack of confidence in keeping my data on the device, once I finally get it set up.
Yours disappointed
Marcos Scriven
Hi Marcos,
Can you confirm what revision firmware your unit is running (I run one here, and I am surprised by some of your comments).
I know there is a new firmware coming out *real soon, and I am sure it will fix many of your quarms.
DR
Hi there - I was running 1.0.2 I think - but thecus support got back to me with this. I've included my reply. To be frank, I'm getting quite annoyed with these Thecus people:
i
Thanks for the response - however, it seems to just raise more for me.
1) Why are you providing links for a mulinational company from <<Strange personal website link>> ? I am very wary to do this - I would expect to see a Thecus.com link
2) Why was this information regarding the LED problem and Snapshot problem not on your website? That's the first place I looked.
3) Regarding the confirmation to cancel - I can assure you I did NOT receive this dialog. Perhaps because I am using Firefox? However, this is now a very popular browser. I am not happy to try this again - particularly considering the rebuild time.
4) Your answer regarding the time to format the RAID array is terrible! Please, could you get one of your engineers to answer this for me? I really fail to believe 1 hour per 250GB disk is acceptable...
5) A lot of your answers are basically suggestions these problems might be fixed in future firmware. However, I am considering returning this product - under Germany statutory distance selling regulations, I have only 7 days, having received the product on Friday.
6) The noise, I am 99% sure, is emanating from the PSU fan. It's much louder than my computer, or any such device I have owned.
7) I don't mean to be rude, but your own grammar in this email is not very good - perhaps this explains the linguistic quality of the Thecus product in general.
Given my disappointment both with the product, and your response, I am attempting to copy this email to the CEO, Edwin Lin. I have tried to guess his email address. If it is wrong, I would appreciate it if you could forward it to him please, copying me in.
Marcos
Support schrieb:
> Hi Marcos,
>
> Thank you for the responses! I agree with you that Thecus should do much better than that. We will do some improvement as soon as possible.
>
>
> 1) Firstly, overall documentation is pretty bad - apart from basic grammatical and spelling errors, it\'s not actually very informative. As one example, out of many, the details regarding "user percentage" is very unclear.
>
> Sorry about that and we will proofread the manual again. And then revise it as soon as possible.
>
>
>
> 2) Even the device itself, on the LCD, has spelling errors. It says 'formattin' instead of 'formatting', even though there are clearly several cells left on the display. Either someone noticed it and couldn't be bother to do anything about it, or simply didn't notice after, presumably, months of development. I am not sure which is worse. This single issue is not a great problem in itself, but I do think it indicates the quality of your products is less than great.
>
> Will correct them as soon as possible! Those wordings could be revised by firmware upgrade later on.
>
>
>
> 3) Here's a big qualm: Why does it take five hours to format 5 250Gb drives? Can this not be done in parallel? If it is done in parallel, that's even more shocking. This is supposedly in 'rebuild speed:high' mode - I shudder to think how long 'low' takes.
>
> We will see if there are any alternatives to speed up the building and formatting process. If possible, later firmware release will improve it.
>
>
>
> 4) Here's the biggie that prompted this rant: Clicking 'remove' on the RAID page does exactly that. What it DOESN'T do is WARN you, or give you a second chance!!! That's it - BOOM, data gone. That sounds like a schoolboy error to me - one of the first rules of software interfaces is to protect the user. Fortunately, it IS what I wanted to do, but it could easily not be in the future, and I'd be VERY wary of even viewing that page, for fear of accidentally clicking the button, or pressing return with the button highlighted... Of course, its taking ANOTHER 5 hours to build the array - surely not... Thought I'd take the time to write this while it's chugging along.
>
> There is a dialogue box asking for confirmation. It says “All disk data will be removed. Are you sure?. However, I thought the default button should be “Cancel” rather than “OK”.
>
>
>
> 5) The reason I chose to remove the RAID array was because there is NO WAY to change the percentages of Snapshot and USB allocations. (By the way, to allow only specification by percentage is plain lazy. Why not absolute values? I think a lot of users would want a less that 1% amount dedicated to USB for instance, say enough for a 4gb card). Why can't you just have a special folder, which can vary in size like any other folder?
>
> Firstly, the partition type of Target USB is different from those of data and snapshot partition. So, once specified, it’s hard to change. Then the snapshot and data partition, there always be potential risks while resizing partitions. We thought data safety is the most important. As for the percentage or absolute value, some users like percentage because it’s easy and simple.
>
>
>
> 6) (This list is getting really quite long for a product I've not even finished setting up) The documentation clearly states that the red LEDs indicate a problem, and yet, from what I can gather from using the N5200, they are also red during formatting, except the top one, which is green. Explaining properly to the user what to expect during different states is again another basic error.
>
> Well, the LED behavior, actually, it should be GREEN to indicate the HDD read/write activity. Please apply attached "ledpatch_1.00.00.2a.rom" to fix it.
>
> [Installation]
>
> 1. Load the patch just like upgrading a firmware. [Admin Web GUI: System / Firmware Upgrade / Browse / ledpatch_1.00.00.2a.rom / Apply]
>
> 2. It requires rebooting to take effect.
>
> During the NAS booting, the first 50 seconds, HDD LEDs may partially glow red or orange. It's a normal behavior and no worries about that please! Once the booting completes, the LED behavior will be: GREEN for HDD activity and RED for HDD error..
>
>
>
> 7) The web interface does not contain any context sensitive help! How easy would it be to put a little question mark icon, or something similar, that one could click to see what that particular item does?
>
> We will think about that.
>
>
>
> 8) The 'snapshot' item in the menu does nothing when I click on it (after I'd setup the RAID array the first time, and the system was reporting 'healthy')
>
> The original firmware v1.00.00.2a doesn’t enable the snapshot function. So, please upgrade to new v1.00.01.3.
>
> [N5200 Firmware v1.00.01.3]
>
> <<Strange personal website link>>
>
> 1. Fixed the problem that RAID creation failed with 750G / 500G drives.
>
> 2. Fixed the problem that writable users could not delete files.
>
> 3. Fixed the problem that snapshot function is gray out in firmware v1.00.00.2a.
>
>
>
> 9) Whomever wrote the content for the web interface does not appear to have had knowledge of the past tense in English. Again, not a great problem, but it's very irritating to see so many errors on a product I have spent so much on.
>
> We will review the content in Web interface.
>
>
>
> 10) The 'status' column on the disk information page contains only the text 'OK?'. I click that to see nothing but 'N/A’ for all the so-called SMART INFO.
>
> It sounds like the disk drive didn’t report SMART info. We did see this symptom on some disk drives.
>
>
>
> 11) Generally, the options and information all seems very limited. For a product of this expense I would expect to see far more information on the component disks, and generally spped and transfer amounts and rates.
>
> Sometimes too many info may confuse NAS users who are not familiar with technical info. But we will consider this suggestion.
>
>
>
> 12) The fan is really rather noisy. There is not need for that level of noise in such a small product. I know there will probably be some stock excuse, but to me in seems like poor engineering, considering the availabilty of quiet fans, and the fact that hard drives don't really need a great deal of cooling.
>
> Could you indicate the noise is from system fan or PSU fan please?
>
>
>
> 13) My main two reasons for purchasing this product, rather than building my own with a Linux distribution, is the small form factor of the case, and the plug and play aspect. In the first aspect, I am pleased - it's a reasonably compact unit, and is well presented. On the later though, I am really very disappointed, Sloppy documention, sloppy web interface, excruciatingly slow RAID setup, and just a general lack of confidence in keeping my data on the device, once I finally get it set up.
>
> Thanks for sharing your thoughts with us. We will review the whole document and do something to make it looks better.
The grammar or linguistic quality in the web UI or documentation is the last thing I care about. Are you going to tell them to spell "colour" instead of "color"? (Just kidding)
I am sure you will get their attention, because I see the same posts on eHomeUpgrade.
As a owner of Thecus N2100, I am sure they will fix the problems as soon as possible, but you have to be patient :).
i can see marcosscriven's point though when he has spent a considderable amount of money on this.
It isn't like the Yes Nano i bought for a few ££ from scan :)
mwk66 - as I said in the post, it's not really that important - but I do feel it's an indicator of quality :)
Unfortunately, the list of things going wrong expands:
1) If you choose "100% data usage", the array builds (after 5 hours), but then it doesn't work. You just get a single line in the logs saying lvm failed to mount. So basically, no matter how big your hard drives or array of drives, you MUST dedicate 1% to the USB read facility. In my case that's 100Mb! That's just a waste of space for me - particularly as the USB read funcationality is pretty flaky as well.
2) The so called 'sanpshot' facility, supposedly fixed in the new firmware they gave me, simply adds a button to make a snapshot. It completely fails to tell you what it's making a snapshot of, or indeed where you might find this snapshot!
The worst part about working around all these bugs is it takes so long to rebuild the array when you change a setting. And then, do you actually trust it with your data? And as for being patient - well, I'm not :) And in any case, I don't think its fair to have to wait for a product to work, I mean with the basic settings bug-free, when you fork out over 800€! Basically, the whole thing completely undermines my confidence in it.
Hi all I currently work for a Design Company who have decided to get hold of some RAID 5 NAS, and from what I can tell this piece of kit seems to do the job well. One thing I was interested in was how well does the Nsync (eew) software work and how does the snapshot system work? Would I be able to attach an external USB HDD and take a snapshot of the whole thing simply by using the front panel?
Ta.
Dave
In my personal and professional experience, raid initialization always takes quite a long time. For instance, on Netapp NAS units, you must first zero all drives, then build the disk array, each operation taking from between 2 to 3 hours to complete. I have also experienced really long build/format times with both software raid configuration and consumer grade hardware raid cards.Quote:
Originally Posted by marcosscriven
RAID init is quite a bit more complicated than a single disk format which is probably why it takes so much longer.
Yes my experience is also that RAID creation can take time.
I have a Thecus 5200 and am currently testing it in a Windows 2003 environment. I am having trouble getting it to recognise Active Directory users and groups. The reviews on both the 5200 and 4100 seemed to gloss over how to set up AD authentication. Has anyone succeeded in getting a 5200 into AD successfully and then using AD users and groups in the ACL screen in the folders option?
Regards
Scott
I just got my Thecus 5200 and am very frustration first because it doesn\'t support AT ALL Barracuda 7200.9 hard drives which just so happened to be the drives I wanted to use in it.
Second, it don\'t authenticate over AFP. public shared folders work find, but as soon as you try to authenticated for an ACL\'s share, nothing works. I had to update the firmware just to get it to enable AFP support and now this. SMB share will work for now, but still sucks I can\'t get AFP support for my mac over the network.