D-Link Forums
The Graveyard - Products No Longer Supported => D-Link Storage => DNS-323 => Topic started by: EBS on April 20, 2011, 07:58:23 AM
-
Hello All,
I am struggling now to get my data off of my DNS-323. I just ordered another one. I hope that when I plug in the new one, my network will spring back to life.
For about two years I gotten used to fast transfers between my two machines and my NAS. I use a lot of large virtual hard disks, and also Acronis disk images. I stash them on the NAS. Transfer was always more than 100 Mb/sec.
Suddenly, four days ago, moving a large file took hours instead of minutes. I am worried that the NAS is failing so I have been moving data off of it..... slowly.
Now when I test with LAN Speed Test (1.1.5 Totusoft), I get 8.1 Mb/sec
Yesterday I replaced all three cables with new cat6. The switch is a Netgear GS605. I have a laptop, a desktop and the DNS-323. The NAS is set to gigabit only. Both computers are gigabit. Transfer laptop to desktop is 110 Mb/sec.
I am recovering family photos now. Transfer is almost complete. Moving 48.5 GB of data has taken just over 14 hours.
My switch tells me that the NAS has a gigabit connection. I have enabled jumbo frames and I have tried various MTU sizes.
The strange thing is that this all ran very differently up until 4 days ago. I did upgrade to 1.09, but I also downgraded as a test two days ago, with no apparent affect. I am back to 1.09 now.
Any ideas?
EBS
-
There are a couple of other things that can impact transfer speed - remember you're moving data from disk to disk, not just across a network - so the time taken to both read from & write to the disks frequently has an impact.
If the disks at either end are fragmented, you will see a drop in throughput.
File sizes also have an impact - transferring 20GB worth of 3MB files takes a heck of a lot longer than transferring 20x3GB files - especially when writing to the DNS-323
-
Anti-virus scanning is often a factor too. Maybe there was an update or you changed some settings?
-
No changes that I can identify. Also, the behavior is the same between the NAS and desktop, and NAS and laptop.
I just completed another test:
A single 307MB file took 28 minutes to copy over. That's 1.4 Mb/sec.
It feels like something is broken. Failing NIC in the NAS?
I will be very curious to find what happens when I plug the same drives into my new DNS-323.
For now many thanks to Wigg and others who posted about [http://www.fs-driver.org/]
I have plugged one of the drives directly into my desktop and I am backing up all data.
-
Did you recently increased the volume of data stored on the unit or has that remained relatively constant before and after the issue surfaced? Is the HDD close to full?
-
I have 2 500GB drives
Volume_1 - RAID 1, ~400MB, 355MB used
Volume_2 - JBOD, ~200MB, 177MB used
I don't think that the volume of data has changed much recently.
-
... Wait, you were getting 100mb/s or 100MB/s? 100 megabit works out to 12.5 megaBYTES.
These boxes were advertised as having transfer rates somewhere between 20 and 25 MB/s.
Do a speedtest using nas tester:
http://www.808.dk/?code-csharp-nas-performance
on a 400MB file I only get 9.55 average write, but a 20.28 average read. This is without jumbo frames.
-
File sizes also have an impact - transferring 20GB worth of 3MB files takes a heck of a lot longer than transferring 20x3GB files - especially when writing to the DNS-323
This one made me smile when I first read it, since read verbatim it is comparing transferring 20GB of small files with 60GB of large files, but in reality that turns out to be the case (at least when it comes to the DNS323) since my experiments suggest that it takes more than six times longer to transfer 20GB of 2MB files compared with 20GB of 2GB files. When I repeat the experiment from PC to PC I see ratios closer to 2.5 times slower, with the speed of the receiving machine (not unsurprisingly) being the main factor in determining the exact ratio.
Just for the record, here are the details of those bits of my setup that participated:
- cheapie gigabit switch (Tenda TEG1008)
- CAT6 cabling to all gigabit devices
- DNS323 (as per my sig)
- Dell GX620 (2.8GHz Pentium D, 3GB RAM, UDMA100 320GB WD HD, gigabit on motherboard identified as Broadcom NetXtreme 57xx in Device Manager, Windows XP 32-bit)
- Toshiba TECRA A10 laptop (2.53GHz Core 2 Duo, 4GB RAM, SATA150 250GB Hitachi HD, Intel 82567LM gigabit on board, Windows 7 Enterprise 32-bit)
- CryoPC Nemesis (i7 2600K overclocked to 4.8GHz, 8GB RAM, SATA300 1TB Samsung F3 HD, Realtek RTL8111E gigabit on motherboard, Windows 7 Pro 64-bit)
NOTE that, other than the NAS, all disk partitions are NTFS.
Current tests were all run on the Dell box (although I intend to repeat them from the Cryo, not in the least because I'm still not convinced that I'm getting the full possibilities talking between Win7 and XP). The data transferred consisted of two flat directories, one with 10 files of 2,146,566,144 bytes and the other with 10240 files of 2,096,232 bytes. The machines at either end of the transfer were not in use for anything else whilst the transfers took place and there was no other significant traffic on the network at the time:
From Dell to NAS, large files took 0:21:18.98 (16.78MB/s) whilst small ones took 2:14:37.14 (2.66MB/s)
From Dell to laptop, large 0:13:57.82 (25.62MB/s), small 0:38:51.27 (9.21MB/s)
From Dell to Cryo, large 0:14:37.69 (24.46MB/s), small 0:37:20.32 (9.58MB/s)
The slower large file transfer speed to the Cryo (compared to the laptop) surprised me somewhat, but I intend to repeat all of these tests a couple of times, so hopefully it will become apparent whether this is normal or a blip.
I also tested locally on those machines that had multiple hard drives:
Cryo (SATA150 Maxtor 160GB as the target), large 0:06:39.32 (53.76MB/s), small 0:08:29.05 (42.17MB/s)
Dell (SATA300 Seagate 160GB), large 0:17:30.98 (20.42MB/s), small 0:26:03.35 (13.73MB/s)
Not sure exactly what conclusions should be drawn from all of the above, and I was quite surprised that the Cryo did such a speedy job with the small files compared to the large, but the one thing that does jump out is that the DNS323 does a far worse job writing small files, as fordem intimated.
-
This one made me smile when I first read it, since read verbatim it is comparing transferring 20GB of small files with 60GB of large files, but in reality that turns out to be the case (at least when it comes to the DNS323) since my experiments suggest that it takes more than six times longer to transfer 20GB of 2MB files compared with 20GB of 2GB files. When I repeat the experiment from PC to PC I see ratios closer to 2.5 times slower, with the speed of the receiving machine (not unsurprisingly) being the main factor in determining the exact ratio.
Steve - that is actually a typo - it's not what I meant to say, but as you discovered, it is not an incorrect statement.
By the way - try defragmenting the drives on the Dell and see what it does for the local transfer.
-
... Wait, you were getting 100mb/s or 100MB/s? 100 megabit works out to 12.5 megaBYTES.
These boxes were advertised as having transfer rates somewhere between 20 and 25 MB/s.
Do a speedtest using nas tester:
http://www.808.dk/?code-csharp-nas-performance
on a 400MB file I only get 9.55 average write, but a 20.28 average read. This is without jumbo frames.
Thank you very much, all of you who are trying to help.
To clarify, In my post, I use [Mb = megabit] and [MB = megabyte]. I hope that this is correct usage. Sorry if it confuses.
Thank you for the link to the NAS speed tester.
My new DNS-323 is here. I have configured it and I have installed a single 500GB drive (Seagate ST3500630AS). The switch tells me that the NAS has a 1000Mb connection.
I have a problem. It must not be with the NAS device. In my first test, I moved a single 102MB file to the NAS using a Windows Explorer copy/paste. It took the transfer 1 hour 45 minutes to complete.
I am testing now with the NAS Performance Tester 1.2.
See log:
NAS performance tester 1.2 http://www.808.dk/?nastester
Running warmup...
Running a 400MB file write on drive N: 5 times...
Iteration 1: 16.05 MB/sec
Iteration 2: 15.78 MB/sec
Iteration 3: 16.04 MB/sec
Iteration 4: 15.95 MB/sec
Iteration 5: 16.43 MB/sec
------------------------------
Average (W): 16.05 MB/sec
------------------------------
Running a 400MB file read on drive N: 5 times...
....... and this is where it still sits after waiting 35 minutes
So writes were very fast in this test (I could certainly live with 16MB/sec). The first of 5 reads has not yet completed.
I believe that during a Windows Explorer copy, the OS both writes and reads the destination drive.
What is my bottleneck? I have disabled my anti-virus on-access scan.
-
NAS performance tester really make it look like I have a read problem:
It finally completed its first 400MB file read:
NAS performance tester 1.2 http://www.808.dk/?nastester
Running warmup...
Running a 400MB file write on drive N: 5 times...
Iteration 1: 16.05 MB/sec
Iteration 2: 15.78 MB/sec
Iteration 3: 16.04 MB/sec
Iteration 4: 15.95 MB/sec
Iteration 5: 16.43 MB/sec
------------------------------
Average (W): 16.05 MB/sec
------------------------------
Running a 400MB file read on drive N: 5 times...
Iteration 1: 0.19 MB/sec
-
I know I've suggested this before - so forgive for being a nag - have you looked at the other side of the connection?
You've swapped out the DNS-323 and pretty much proven that it's not the problem.
Have you tried defragmenting the disk(s) in the PC, have you tried a new network cable between the router & the PC, have you looked at the NIC and/or it's drivers?
-
I am seeing average write speed 16.79MB/s and read speed 18.97MB/s. That's still fairly slow for a gigabit connection using a single 750GB disk - no raid.
I'll try changing network cable and try it again.
-
I am seeing average write speed 16.79MB/s and read speed 18.97MB/s. That's still fairly slow for a gigabit connection using a single 750GB disk - no raid.
I'll try changing network cable and try it again.
What you need to understand is that anything over 100mbps (call it 10MB/sec to account for overhead and make it convenient) is gigabit - and - not because you have a pipe capable of flowing 1000mbps, means you'll see anywhere close to that.
Changing the cable probably won't help.
-
I am getting over 50MB/s to another server on the same network easily. Call it what you want :-)
-
I am getting over 50MB/s to another server on the same network easily
Two questions to ask yourself. How much did that other server cost in comparison to the DNS323?? What is the power footprint of that other server compared to the DNS323??
FWIW I doubt (based on my own testing and what I've read here and elsewhere about the device) that you will do any better than you are now, whatever else you change in your infrastructure. After a quick perusal I can't find anything in the specifications on the web site, or the data sheet, but the manual contains this statement:
"High Performance Gigabit Ethernet Connectivity (Up to 23/15MBps or 184/120Mbps Read/Write)"
which strikes me as a fair reflection. I regularly see write speeds of 15MB/s (and indeed a little better than that, up to about 17.5) and the best read speed I've managed is 23.7MB/s (although I've not done anything like as much read testing - most of my usage of the NAS is as a backup of backups, hence I tend to be writing large files). Depending on the number of files you are handling (in relation to the total amount of data) these numbers can show a downward trend pretty quickly, especially when writing.
-
Hi Steve,
Uf, who reads manuals. Now, if this is true there are some consequences:
- it is a misleading advertisement - nowhere on specs this limitation is listed
- it is a ****py port or hw (who knows where does it origin)
- it is a waste of time
- hey, it is a D-Link!
Hey, it is just a throughput, nothing special: get data from lan and send it to disk and vice versa. Why would it slow down is beyond me.
PS. That server is a cheap homebuilt one with old hw, motherboard lan and a cheap disk
-
who reads manuals
Well, me obviously
it is a misleading advertisement - nowhere on specs this limitation is listed
So not misleading at all, merely incomplete. The fact that nothing much is made of the speed of the unit really ought to tell you that this isn't being sold as a top-end unit (as ought the price)
it is a waste of time
I would disagree with that sentiment. I purchased the unit for less than sixty quid, working to a budget, and it gives me exactly what I was expecting for the price - a cheap NAS. Anything with better performance was going to cost me at least £150 more, so I certainly don't view the unit as a waste of anything. Your mileage obviously varies
Why would it slow down is beyond me
Because the processor in the DNS323 simply isn't capable of handling the data any faster (with the supplied firmware, whether it would be possible to improve things with some hand-crafted assembler code is a different matter)
-
OK, some statements were rhetorical. But seriously, a hw incapable of handling 50MB/s transfer. Watch, not data manipulation, just mere throughput? Seriously?
If the controller on the disk can do it and a $10 network card can do it you would seriously think that a modern CPU can't handle it? Obviously there is a some sort of bottleneck in there and this is hugely disappointing even for a cheap device like DNS-323 is. But that's subjective.
As per "So not misleading at all, merely incomplete" I tend to disagree. What would you think if you'd buy a 10TB hard disk with 1GB/s speed capability (advertised) only to find out that it really can store only 1GB of data at 1Kb/s rate? Incomplete? Misleading? Fraud? Pick one.
-
As per "So not misleading at all, merely incomplete" I tend to disagree. What would you think if you'd buy a 10TB hard disk with 1GB/s speed capability (advertised) only to find out that it really can store only 1GB of data at 1Kb/s rate? Incomplete? Misleading? Fraud? Pick one.
Does it not say on the outside of the box "upto 23MB/sec" (or something similar - I forget the exact number, it's been four years since I've seen the box mine came in - but the throughput numbers ARE on the box) - isn't the real problem that you failed to research the suitability of the device to meet your needs?
For the sake of discussion ...
I have behind me a pair of almost identical IBM xSeries small business servers - they are 2006 technology - 3GB RAM, 3GHz Pentium IV, integrated Intel PRO/1000 NIC - they both have 250GB SATA disks, one has 2x250GB in a RAID1 configuration on the integrated ServeRAID 7e controller, the other has 3x250GB in a RAID5 on a ServeRAID 7t controller.
The first system is capable of almost 800mb/s continuous transfers, the second will start out at 800mb/sec and in a matter of seconds drop to around 80mb/sec - do you feel that there is something, incomplete, fradulent or misleading about this?
-
Does it not say on the outside of the box "upto 23MB/sec" (or something similar - I forget the exact number, it's been four years since I've seen the box mine came in - but the throughput numbers ARE on the box) - isn't the real problem that you failed to research the suitability of the device to meet your needs?
I forgot what box looks like as you did. The one problem with the box is that you actually can't see one when buying over internet, can you - and be assured no merchant is listing this limitation? Anyway when I usually buy something over internet I look at specifications on the manufacturer web site. Why is this limitation missing there is beyond me. Granted I could research more (reviews, etc.) but honestly, didn't even think about it could so severely limit transfer speed.
For the sake of discussion ...
I have behind me a pair of almost identical IBM xSeries small business servers - they are 2006 technology - 3GB RAM, 3GHz Pentium IV, integrated Intel PRO/1000 NIC - they both have 250GB SATA disks, one has 2x250GB in a RAID1 configuration on the integrated ServeRAID 7e controller, the other has 3x250GB in a RAID5 on a ServeRAID 7t controller.
The first system is capable of almost 800mb/s continuous transfers, the second will start out at 800mb/sec and in a matter of seconds drop to around 80mb/sec - do you feel that there is something, incomplete, fradulent or misleading about this?
No, not at all. RAID 5 is slow and 80MB/s sustained speed is in line with what one would expect (read should be faster). These are physical limitations of RAID 5 and works as expected. Now, if the first one would perform like the second one it would be odd.
The thing is that you are talking server configuration where you can achieve full speed (first one) and you can slow it down with your "mistakes".
-
Might as well jump in here:
As per "So not misleading at all, merely incomplete" I tend to disagree. What would you think if you'd buy a 10TB hard disk with 1GB/s speed capability (advertised) only to find out that it really can store only 1GB of data at 1Kb/s rate? Incomplete? Misleading? Fraud? Pick one.
Has anyone on the planet ever gotten the advertised 3 gb/s or 6gb/s of a SATA drive? Actually ever use a 1TB HDD for that matter? Reserve space for sector reallocation/system use/over head aside, nobody gets a TB. Ever see an ad for McDonald's for that matter?
So you didn't know the limits of the box before you bought it? Caveat emptor. The specs are on the data sheet ON the dlink website:
http://www.dlink.ca/products/?pid=509
ftp://ftp10.dlink.com/pdfs/products/DNS-323/DNS-323_ds_ca.pdf
Google "dns-323 performance" or "dns-323 transfer speed" there are tons of reviews!
When your box arrives and you see the specs (which are written on the box) you don't open the damned thing! You return it. End of story.
-
Might as well jump in here:
Has anyone on the planet ever gotten the advertised 3 gb/s or 6gb/s of a SATA drive? Actually ever use a 1TB HDD for that matter? Reserve space for sector reallocation/system use/over head aside, nobody gets a TB. Ever see an ad for McDonald's for that matter?
So you didn't know the limits of the box before you bought it? Caveat emptor. The specs are on the data sheet ON the dlink website:
http://www.dlink.ca/products/?pid=509
ftp://ftp10.dlink.com/pdfs/products/DNS-323/DNS-323_ds_ca.pdf
Google "dns-323 performance" or "dns-323 transfer speed" there are tons of reviews!
When your box arrives and you see the specs (which are written on the box) you don't open the damned thing! You return it. End of story.
AMEN!!!!!!!!!!!!
-
Has anyone on the planet ever gotten the advertised 3 gb/s or 6gb/s of a SATA drive?
Yes, it is possible for reading/writing out/to the onboard cache or using modern SSDs. Is it possible to achieve 1Gb/s out of DNS-323?
But again these aren't good comparisons. Disks have known limitations, don't they?
Actually ever use a 1TB HDD for that matter? Reserve space for sector reallocation/system use/over head aside, nobody gets a TB. Ever see an ad for McDonald's for that matter?
Again, these are well known limits.
So you didn't know the limits of the box before you bought it? Caveat emptor. The specs are on the data sheet ON the dlink website:
http://www.dlink.ca/products/?pid=509
ftp://ftp10.dlink.com/pdfs/products/DNS-323/DNS-323_ds_ca.pdf
Yeah, well, pardon me for not reading every variant of specifications available. But that's not about me, it is about how D-Link present specs.
Google "dns-323 performance" or "dns-323 transfer speed" there are tons of reviews!
When your box arrives and you see the specs (which are written on the box) you don't open the damned thing! You return it. End of story.
Again, this isn't about being carefully (in the sense of scouting the web before) when buying products, but rather about misleading specifications.
And here I am asking again you wise men protectors of D-Link. Why is the speed that low? I have yet to see an explanation. And don't start again with "you have to google before buying".
-
Hold a second - you're going to excuse the inability of a disk drive because of well know limits but have a problem with the available network bandwidth on the DNS-323's gigabit port - do I have it right?
Would you be happier if I said I have pushed the ethernet port on a DNS-323 to 400mbps or 50MB/sec - and yes I am 100% serious about that - and let's keep this in perspective - I have other gigabit NICs (realtek chipsets) that cannot achieve anything beyond 270 mbps.
This is not about how D-Link presents specifications or misleading specifications, this is about incorrect expectations - regardless of what you expect the ethernet port meets IEEE standards for 802.3ab.
-
this isn't about being carefully (in the sense of scouting the web before) when buying products, but rather about misleading specifications
I'm really not sure what you are expecting from this supposedly misleading information. Having taken a look at the specs for similar devices from Seagate, Netgear and Buffalo I don't see any of their specifications trumpeting their transfer speeds either. Even QNAP or Thecus don't list transfer speeds in all of their spec sheets. Some of the more expensive units feature phrases like 'up to 60MB/s' in their feature descriptions but the entry level and low-end devices are universally silent on the subject.
Why is the speed that low? I have yet to see an explanation
Que?? What was this then??
Because the processor in the DNS323 simply isn't capable of handling the data any faster (with the supplied firmware, whether it would be possible to improve things with some hand-crafted assembler code is a different matter)
-
Steve, procesor not being able to transfer speeds of above 24MB/s. Are you really serious? Besides transfer isn't done by CPU anymore or the unit has a Z80.
Besides, you are saying CPU, fordem is saying network interface. Clear? Sure.
-
Besides, you are saying CPU, fordem is saying network interface. Clear? Sure.
I said what???
-
I said what???
"Would you be happier if I said I have pushed the ethernet port on a DNS-323 to 400mbps or 50MB/sec - and yes I am 100% serious about that"
This is not about how D-Link presents specifications or misleading specifications, this is about incorrect expectations - regardless of what you expect the ethernet port meets IEEE standards for 802.3ab.
Perhaps I've misunderstood you that onboard network port is the bottleneck?
-
You're griping about <20MB/sec throughput, I make a statement about pushing 50MB/sec through the NIC and you somehow infer that I'm saying the NIC is the bottleneck?
Doesn't that sound backward to you?
-
You're griping about <20MB/sec throughput, I make a statement about pushing 50MB/sec through the NIC and you somehow infer that I'm saying the NIC is the bottleneck?
Doesn't that sound backward to you?
Ah, ok, I've thought you've hacked it to perform better. My bad. So you are saying that you pushed it to 50Mb/s with what? Dummy data?
-
procesor not being able to transfer speeds of above 24MB/s. Are you really serious?
Yes. It isn't just about pushing around data, it is everything from managing the IP stack to checking directory structures, permissions etc. I have been doing a fair bit of testing both of the DNS-323 and my network and I've not managed to better 24MB/s when copying various large workloads between the two oldest machines on my network - two five year old desktops, one a Dell GX620 with a 2.8GHz Pentium D and the other a Mesh (UK box shifter) Athlon 3500+, both with more RAM than the NAS - which are of a similar age to the original release of the DNS-323.
I know that my network infrastructure is capable of better than that because I can get 40MB/s or better when shifting data between my new CryoPC i7 2600K @ 4.8GHz and the two and a half year old Toshiba laptop with a 2.53GHz Core 2 Duo (and could no doubt do a lot better than that if I had two i7 beasties and involved the SSD rather than limiting myself to the HD). None of my machines are optimised for a file server role, and could probably be tweaked to perform better than they do, but there are all more powerful than the processor in the NAS and would seem to provide comparable data points.
-
Ah, ok, I've thought you've hacked it to perform better. My bad. So you are saying that you pushed it to 50Mb/s with what? Dummy data?
Rev A hardware, only "mod" is a custom fun_plug that loads a telnet daemon so that I have shell access.
-
Yes. It isn't just about pushing around data, it is everything from managing the IP stack to checking directory structures, permissions etc.
Sure, there are some checks to be performed. But once those are done it boils down to transfer from network to disk (imagine big files) or in other direction. If I were to build a (cheap) NAS I would take care of transfer speeds and robust firmware first and then I'd add a ton of other more or less usable features. IOW even when it comes to a cheap NAS like this is, transfer speed should depend on disks only, not on anything else. Perhaps the speed is limited by lousy firmware and not hardware.
-
Hello all,
Since I started all this, I thought I should post the resolution.
It was the switch.
I replaced the switch and I get 55Mbps both to and from the 323(s). I can plug the old switch back in and duplicate the problem. The only symptom is that transfers from either 323 through the switch drop down to 8Mbps or less (often 1Mb).
EBS
-
Hello all,
Since I started all this, I thought I should post the resolution.
It was the switch.
I replaced the switch and I get 55Mbps both to and from the 323(s). I can plug the old switch back in and duplicate the problem. The only symptom is that transfers from either 323 through the switch drop down to 8Mbps or less (often 1Mb).
EBS
Hi there,
Just been following this thread I also feel I'm suffering from pretty low transfer speeds. I'm new to using a NAS, so when you say that you 'changed the switch', what does that mean?
Thanks alot,
Tom
-
I also feel I'm suffering from pretty low transfer speeds
Any concrete data?? How slow is 'pretty low'??
when you say that you 'changed the switch', what does that mean?
Nothing to do with the NAS as such, the switch in this context is a network switch (a device that routes data from one directly attached device to another). Many home networks don't use switches but instead may have a network hub (which serves the same purpose as a switch, but broadcasts any data to all directly attached devices, and becomes considerably less effective as the number of directly attached devices increases) or a router of some description (which can route data to different networks and will normally be connected to both the Internet and the local network in a SOHO type environment). These days many home setups just have a broadband router which, in addition to connecting to the outside world and providing a handful of wired network ports, will also have a wireless capability.
Bottom line is that the original problem was nothing to do with the DNS323 but instead was probably a fault in the network switch (a Netgear GS605, as per the OP).
Moral of the tale is probably that selling NAS devices as something that you can simply plug in and get working well without any technical knowledge of networking is probably a little optimistic and no doubt results in a significant percentage of the support requests for these kinds of devices.