D-Link Forums
The Graveyard - Products No Longer Supported => D-Link Storage => DNS-323 => Topic started by: gunrunnerjohn on August 05, 2010, 01:16:49 PM
-
Here's my latest round of NAS tests, you can clearly see why I prefer the Synology. :)
It's interesting how much performance difference there is between the DNS-321 and DNS-323, I thought it was less before...
Synology DS209 Gigabit, no jumbo frames, RAID-1 EXT3 1.5TB Seagate 5400 RPM Drives
Running a 2000MB file write on drive I: once...
------------------------------
Average (W): 31.87 MB/sec
------------------------------
Running a 2000MB file read on drive I: once...
------------------------------
Average (R): 62.97 MB/sec
------------------------------
DNS-323 Gigabit, no jumbo frames, single EXT3 1.5TB Hitachi 5400 RPM Drive
Running a 2000MB file write on drive j: once...
------------------------------
Average (W): 13.03 MB/sec
------------------------------
Running a 2000MB file read on drive j: once...
------------------------------
Average (R): 17.3 MB/sec
------------------------------
DNS-321 Gigabit, no jumbo frames, single EXT3 1.5TB Hitachi 5400 RPM Drive
Running a 2000MB file write on drive K: once...
------------------------------
Average (W): 9.34 MB/sec
------------------------------
Running a 2000MB file read on drive K: once...
------------------------------
Average (R): 11.81 MB/sec
------------------------------
-
Wow thats a lot of difference between the Synology and the DNS 323. I wish I had seen these posts before I bought my DNS 323. :'(
-
They're in a different class, the Synology is my prime NAS now, I use the others for unattended backups where the speed isn't an issue.
-
That figure you have for the read on the 321 is the exact amount for a 100M ethernet cutoiff. I had the same problem until I manually set all devices on my system to gigabit ethernet. Previously they had been set to auto negotiation and reported 1gig buit they were not running at 1 gig. The figures you have are too low for the 323.
On another point, I was doing a lot of drive shifting since yesterday (about 3gig worth of stuff!) and my linux machine with sata drives was only maxing out at 50MB/sec. That's internal moving from one eco drive to another eco drive.
I definitely think the time has come for the next generation data transport - 1gig cards are just way too much for what most computers are capable of in a real world scenario.
-
The DNS-321 is indeed on a gigabit switch, and has end-to-end gigabit connectivity.
Speed 1000 Mbps
-
Yea - mine said that too, but it wasn't. I had auto set and the readout said 1000. Also, the detail section of the network readout for my NIC card said 1000. But it wasn't - both readouts were lying ;)
If you go to your NIC settings, set to Gigabit only, then take the DNS off auto and set to 1000 in the setting instead, I reckon you will find a big difference.
Or, if you are convinced you are right then no need to bother really ;)
-
Well, since the NIC that is used to benchmark that NAS is the same one that did the test to the Synology box, it's pretty hard to believe it's running at 100mbit. If it is, I'd sure like to hear you explain how that works. :)
I manually configured the DNS-321 for gigabit only and ran the same test. I also checked the switch connect indications to insure that a gigabit connection was being made by the DNS-321 and the feed into the switch. Finally, due to the fact that some switches are known to have problems mixing 100mbit and gigabit traffic, I disconnected the NDAS 100mbit drives also on the switch, leaving only the incoming gigabit feed and the DNS-321. A transfer to my laptop on the same switch managed to hit 40mbytes/sec, surely indicating the switch was actually receiving and handling gigabit traffic.
With that configuration, I tried your little experiment, and I got the same result.
The results:
Running warmup...
Running a 2000MB file write on drive N: once...
------------------------------
Average (W): 9.62 MB/sec
------------------------------
Running a 2000MB file read on drive N: once...
------------------------------
Average (R): 13.45 MB/sec
------------------------------
Or, if you are convinced you are right then no need to bother really
Any more "pearls of wisdom" you care to throw my way? Perhaps you shouldn't assume everyone but you is incompetent. ::)
-
I have my finger on the trigger of a synology DS210+ I like the IP camera feature, and it looks like their software doesnt suck.
-
I've had a few issues with the Synology software as well, the USB backup was problematic until I get a special version of patched firmware. For some brands of external USB drives, the backup still has issues.
The principal reason I got it was for the network speed. :)
-
syonology does 4k sectors?
-
@tank-killer: If you look at this link http://www.synology.com/support/hd.php?lang=us and put in the model (210+) it shows support for the 4k sectors.
@grj: Like you I finally gave up on D-Link so I ordered a DS210+. I should get it in a few days. I was going to get a DS209 but I got this DS210+ for about the same money. Thanks for the advice.
For as much time as I have put into trying to make the dns-323 work it has not been worth the difference in price between the DS209 and the dns-323.
Terry
-
syonology does 4k sectors?
I picked my model, the DS209, and checked the drive compatibility list, here it is: http://www.synology.com/support/hd.php?lang=us&product_id=29 (http://www.synology.com/support/hd.php?lang=us&product_id=29)
-
Like you I finally gave up on D-Link so I ordered a DS210+. I should get it in a few days. I was going to get a DS209 but I got this DS210+ for about the same money. Thanks for the advice.
For as much time as I have put into trying to make the dns-323 work it has not been worth the difference in price between the DS209 and the dns-323.
Well, it's really somewhat of an unfair comparison. The DS209 cost twice what the DNS-323 costs, so I would expect it to offer some substantial performance improvements. However, if the average performance that most folks get with the DNS-323 isn't sufficient, it's a good move. :)
-
Do you not get tempted though at the upper price rang to have a go at building your own? I've been doing a lot of linux building lately and been looking at some neat cases and HTPC parts and am thinking that my next project will be to build a NAS. I think I can do it cheaper than one of the premium NAS's and faster. The main thing will be the case - it has to look cool and be small :) I might even see if I can talk someone into building me a custom case...
-
I wasn't tempted at all. If you consider the true cost of ownership, the Synology comes out ahead of anything I could build. It's not only the cost of the NAS, but also the power to run a real Linux box. Building a custom case will most certainly cost more than the Synology all by itself! :o
I have in the past had a PC running headless for my file/print server, but I find the NAS to be much more efficient in terms of size and power.
I don't doubt I could put something together that would outperform most NAS boxes, but I really have no need for that kind of performance. The Synology does all I need and more, and I typed into an order form and the completed product arrived at my doorstop in a few days for $300 total. ;) All I had to do is stick the disks into the box and run the setup.
-
I'm glad that the forum seems active and people are reporting speed issues.
I recently bought the DNS-321 and could get at most 5-7MB/s over gigabit ethernet. I've tried SMB and FTP, EXT2 and EXT3, Raid 0, Raid 0, JBOD, Firmware 1.2 and 1.3 ... it's all been bad. I've even run a crossover between the NAS and my PC so no switch was involved. I've been extremely disappointed and even more so with the e-mail support. 6 e-mails back and forth and the only suggestion I've gotten was to reset my router - LAME.
What I don't understand is that smallnetbuilder has both the DNS-323 and DNS-321 at about 15MB/s http://www.smallnetbuilder.com/nas/nas-reviews/30521-move-over-dns-323-d-link-dns-321-reviewed?showall=&start=3 (http://www.smallnetbuilder.com/nas/nas-reviews/30521-move-over-dns-323-d-link-dns-321-reviewed?showall=&start=3) with a unit that DLink supplied. What gives for the rest of us?
I had an expectation that performance would be around what was in the review. Clearly it is not and DLink won't support the device. Had I known that transfers would be less than half of what the reviewers got I would have gone with another product or just kept my own home built box.
-
Jumbo frames?
-
For reasons I'll never understand, I can't seem to get any benefit from jumbo frames, they actually slow things down. I know something must have been added to the network to affect things, because they used to give me about a 20% boost. I've tinkered and tinkered, and jumbo frames slow all my boxes down, including the Synology. I even bought a new 8 port gigabit switch in hopes that something about the switch was the issue, no sale. The packets seem to get through, and I can do ping tests end to end and verify that jumbo packets are capable of being passed, but they still slow things down. I have tried it from three different hosts, same issue.
If I ever get time, I may go after this again, but it's clear that something is amiss in jumbo-land. :)
-
I've tried jumbo frames and haven't had any success either.
One thing to be aware of is that all frame sizes aren't necessarily supported by your card. I have one of the Intel Pro/1000 cards and it only supports 9K frames. So be careful and make sure to match your frame size with something your card can support.
-
Well, the really odd thing was that at one time jumbo frames made about a 10-15% improvement. I have the same machines, the same set of gigabit switches, and the same NAS, and now they reduce my performance! I was using 4k jumbo frames, but I tried 4k and 9k with the same results.
Obviously, something changed in the network that affects it, but I'll be damned if I know what it was. Pings across the infrastructure with a 9k packet work, so I know the switches are passing the jumbo packets.
-
I found 10% increase with jumbo frames. I found a variety of things affect jframe affecttiveness.
Flow control in a mixed 100/1000 enviroment, shutting it off in the switch AND client seemed to improve affectiveness.
How much broadcast traffic you have on your network (especially with flow control enabled)
Length of cable run to the client (dunno why but i get better performance from a longer cable)?
The quality of your switch, I have had a regular gb switch (that passed 9k frames) that i got poor performance from. However after i replaced it with my lvl2 managed switch (which allows me to control broadcast traffic and flow control) jframes helped alot.
Clients NIC quality + drivers
The size of files you are transferring, as with any transfer. Its more affective on 50mb+ files.
-
Hard to know what the difference is.
-
I have 2 Seagate 1TB drivers (ST31000333AS) in my DNS-323 in Raid1 configuration. DNS-323 has jumbo frames enabled and MTU set to 9000. DNS is connected to a Linksys SRW2008 switch, which has jumbo frames enabled. Spanning Tree is also enabled in the switch, but don't think it would affect the performance in any way. I also have a computer Vista 32bit Pro with 10/100/1000 Atheros AR8121/AR8113/AR8114 NIC connected to the same switch. Trying to copy a 2GB file from DNS to the internal HDD of the workstation gets me 9-10mb/s, which is somewhere around 70Mb a sec.
Does anybody know what can be done to get speeds close to 15mb mentioned in the 1st post of this thread?
Thanks
P.S. DNS is not being accessed by anyone else when I copy files, nor does it have anything scheduled for downloading under applications.
-
For the DNS-323, did you try disabling jumbo frames. That's what the configuration is in that post you reference.
-
Just thinking out loud here, but if you adjust the MTU size on the DNS-323, shouldn't you also adjust it on your network interface card to match, otherwise you'll get lots of fragmented packets and performance loss. Wireshark or similar should identify that as the cause.
You can change it on the cmd line using a command such as
netsh interface ipv4 set subinterface "Network Connection" mtu=xxxx store=persistent
-
Obviously, you have to have all the MTU's matching in the system. :)
-
How is the DS210+ different from the DS210j?
-
How is the DS210+ different from the DS210j?
Spelling
.. and other things
http://www.synology.com/enu/products/DS210j/spec.php (http://www.synology.com/enu/products/DS210j/spec.php)
http://www.synology.com/enu/products/DS210+/spec.php (http://www.synology.com/enu/products/DS210+/spec.php)
-
Obviously, you have to have all the MTU's matching in the system. :)
How does that work, when your ISP tells you to make sure the router's MTU is the same as your computer's MTU? And the ISP usually has a specific MTU they want you to use.
-
The router's WAN MTU has nothing to do with your network and jumbo frames.
-
beefy324:
The ds210j runs about half the data throughput as the ds210+. It also does not have an esata port. You might also look at the DS209.
Terry
-
The "J" models are lower end units, I'd stick with either the standard ones or the + models.
-
The router's WAN MTU has nothing to do with your network and jumbo frames.
I read that you should set the MTU of the NIC to match that of the WAN's MTU :-\
http://www.talktalkmembers.com/forums/showthread.php?t=41131 (http://www.talktalkmembers.com/forums/showthread.php?t=41131)
Although you may have now entered the right MTU setting in the router, you may need to adjust the MTU in your computer system, as this may still be showing a default MTU which is different from your router MTU
Confused.
-
Confuses me as well. :) I've always been on cable or fiber, and I've never varied from the defaults. I believe there is a setting for the gateway MTU as well as the jumbo frames. Clearly, a frame of 9000 exceeds the 1492 or 1500 that is the general sizes for DSL and cable. Truthfully, I never gave this issue much thought, since it's never bitten me in the butt. :D
-
beefy324:
The ds210j runs about half the data throughput as the ds210+. It also does not have an esata port. You might also look at the DS209.
Terry
The "J" models are lower end units, I'd stick with either the standard ones or the + models.
Thanks!
But for the price of $199, it seems to be of better value compared to the DNS 323. The average write performance is around 50+ MB/s compared to the 20s MB/s of the Dlink and the read performance is around 55MB/s
-
Yep, the DS210+ would probably be a better choice. I'm sure it'll far exceed the DNS-323 in performance.
-
I ordered the 210+ once I pair it with the right HD's I will post my results here
-
Pretty much any decent 5400 or 7200 RPM drive will do the trick.
-
Thanks!
But for the price of $199, it seems to be of better value compared to the DNS 323. The average write performance is around 50+ MB/s compared to the 20s MB/s of the Dlink and the read performance is around 55MB/s
You are correct. The DS210j would be a better choice than the DNS-323. The firmware works and they upgrade it on a regular basis. Even though it is not as fast as the DS209 or the DS210+ it still should be pretty quick and since it appears that they use the same base firmware on all models from the end user standpoint you would get the same features as the rest only not as fast.
Terry
-
You are correct. The DS210j would be a better choice than the DNS-323. The firmware works and they upgrade it on a regular basis. Even though it is not as fast as the DS209 or the DS210+ it still should be pretty quick and since it appears that they use the same base firmware on all models from the end user standpoint you would get the same features as the rest only not as fast.
Terry
Are there any issues with synology? I plan to use it with my NMT(popcorn Hour) via NFS. Currently, I have no problems with my DNS 323 streaming bluray ISOs or BDMN folders. However, I experience stuttering when I use 2 media players and use the DNS to stream content.
Also, can transmission be installed with this synology?
-
Also, can transmission be installed with this synology?
Apparently, yes: http://forum.synology.com/enu/viewtopic.php?f=3&t=17403
-
Actually, I just turned Jumbo frames off and now I have gained another 3.5 MB/second
-
I was able to get ~13MB/s with 9K jumbo frames over FTP on a DNS-321.
SMB was a lot slower <6MB/s. I've narrowed it down to a bug in the 3.028 version of samba.
I'll have to figure out have to get a later version installed. Maybe against the ffp toolchain.
-
Have you tried NFS rather than SMB?
-
I was able to fix the problem on my own. The performance problems without a doubt are related to the specific version of samba that comes with the default firmware. The firmware samba version is 3.0.28. I installed the ffp version of samba version 3.3.2. The difference is night and day. Transfers went from 4-5MB/s to >8MB/s on 10/100 ethernet. That's a 50-60% improvement. I wish DLink would get on the ball and update the official firmware with the updated software and add telnet and rsync while they're at it. Their support has been beyond worthless.
Here's my ffp samba section:
echo "* Disabling samba 3.0.28 ..."
smb stop
echo "* Loading samba 3.3.2 ..."
/ffp/sbin/smbd -s /etc/samba/smb.conf --smb-passwd-file=/etc/samba/smbpasswd
echo "* OK"
This allows you to use the configuration file from the web. However, I'm pretty sure you have to reboot to get it to take effect since the web config launches the old config script. I'm sure you could modify the "smb" script to fix this as well.
-
It's strange that you guys are seeing performance issues.
My DNS-323 is connected to a 10/100 router and my Windows 7 laptop is Wireless N 300Mbps.
I've just transferred a 2.5GB file to the DNS-323 and Windows reports a speed of ~6.28-7.24MB/second without modifying anything.
The same test connected via an ethernet cable to the router shows ~10MB/second
-
I'm on a DNS-321 with Firmware 1.3. I suspect the DNS-323 has a different samba version than the 321. Could you report back with your DNS-323 firmware version and the samba version.
-
F/W: 1.08
Samba: 3.0.24
-
Thanks. This is consistent with my research. In version 3.0.28 they added some additional "chattiness" to resolve some other bug which decreased performance. The slowless was resolved in a later version of samba. This is also consistent with why smallnetbuilder saw much better performance using the 1.0 version of the DNS-321 firmware. On a device with so little CPU power this was amplified. One thing I do notice is the the CPU usage during a samba transfer hovers around 50-80% so there is probably another bottleneck in the system.
-
It's interesting to note that I think my fastest times were with firmware 1.07...