Need some pointers on improving network Speed Please

[leeland]

I recently upgraded my Home Network (1 PC, TS-209 NAS, & Xbox 360) to GB network.

I connected all devices with CAT 5e and am using a DLINK-DGL4100 GB Gaming Router.

I have a fourth device ( a wireless Belikin router hooked up via CAT 5 cable to the last port on the router to feed my wireless connection in the house & vonage VOIP Phone router)

I installed Vista on my PC with SP1 installed and was all ready to try out some file copies (Large # of Small files) and (Small # of Large Files) to get an idea of the speed improvements...if any over my previous 100 mb router.

The MAX/PEAK speed I reached was 13 mb's/sec when copying a 6 gb .ISO file from my desktop TO the NAS

The Normal speeds I am seeing are in the 3 - 9 mb's/sec range...

I am no network guru by any means so I am not sure what the expected MAX throughput is on a GB network with all devices on CAT 5e or better cables...however I think it should be better than what I am getting.

What I have done so far
1. The NIC on the PC is a GB NIC and it is showing up as a GB connection in the Network Properties...

2. I set the Jumbo Frame Setting MTU on the NAS to ENABLE jumbo frams and the MTU value is 9000 bytes

3. I "attempted" to configure the same value on the NIC properties but didn't see any improvment...

QUESTIONS
1. Can anyone give me some recommendations or improving the speed (Any specific settings to review?)
2. Also can someone give me what the expected Normal speeds you would expect to see on a GB HOME network?


Thanks in advance,

Lee

 

[Madwand1]

According to SmallNetBuilder's NAS charts, the TS-209 Pro tops out around 17 MB/s sustained RAID 1 writes for large files in a synthetic test. As such a test factors out the source drive speed, actual file transfers could be slower in practice, so your 13 MB/s might not be far off the expected performance of this device.

You could try 4K frame size, as this has been found to sometimes work better than 9K frames, and SNB's results were based on 4K.

http://www.smallnetbuilder.com...omposite_write_big.png

Consumer NAS boxes often have such performance limitations due to hardware compromises. For comparison 30 MB/s is a commonly-achieved transfer rate for Windows to Windows over GbE. Of course this varies widely; it's not achieved in some cases, and significantly exceeded in others.

Edit: AFAIK, the DGL-4100 line doesn't support jumbo frames at all, so 9K or 4K jumbo frames shouldn't work. Perhaps jumbo frames are misconfigured and thereby not coming into effect and then breaking the transfers? If you're not using a separate jumbo-capable switch, you should simply disable jumbo frames and then use SNB's non-jumbo results for comparison.

 

[Markbnj]

Agree with Madwand1. I get about 35 MB/s drive-to-drive (windows to debian) across a gigabit LAN using a DIR-655. Some folks here suggested iperf, which is a great tool for measuring actual network throughput. If you want to know how the network is performing you should get the NAS out of the equation.

 

[NickOlsen8390]

The other day copying a Linux iso from my server to my desktop i got 64MB/s. Both are GigE with Jumbo frames. Server is a AMD 5000+ with 2gb of ram. And the drive was was copying from was one of the new 7200.11 seagate 500gb drives. And was copying to my raid 0 array of 2x 250gb seagate 7200.10 drives. Also, server is XP pro, and desktop is vista ultimate X64 SP1. Switch is a HP Procurve 1800-24G


And as for the OP, i would look at the NAS, Try it from another computer and see what you get. I have a small D-link nas that is on GigE and it only moves like 8MB/s

 

[leeland]

Originally posted by: Madwand1
According to SmallNetBuilder's NAS charts, the TS-209 Pro tops out around 17 MB/s sustained RAID 1 writes for large files in a synthetic test. As such a test factors out the source drive speed, actual file transfers could be slower in practice, so your 13 MB/s might not be far off the expected performance of this device.

You could try 4K frame size, as this has been found to sometimes work better than 9K frames, and SNB's results were based on 4K.

http://www.smallnetbuilder.com...omposite_write_big.png

Consumer NAS boxes often have such performance limitations due to hardware compromises. For comparison 30 MB/s is a commonly-achieved transfer rate for Windows to Windows over GbE. Of course this varies widely; it's not achieved in some cases, and significantly exceeded in others.

Edit: AFAIK, the DGL-4100 line doesn't support jumbo frames at all, so 9K or 4K jumbo frames shouldn't work. Perhaps jumbo frames are misconfigured and thereby not coming into effect and then breaking the transfers? If you're not using a separate jumbo-capable switch, you should simply disable jumbo frames and then use SNB's non-jumbo results for comparison.

Thanks for the excellent post Madwand1...that would make sense...I will change things back to the default and see what kind of speeds I get...

In the event I was able to get a router/switch that did support JUMBO frames how much of an impact would that have as far as improving the speed

vs.

A standard router/switch that didn't support JUMBO frames?


Thanks again,

Lee

 

[drebo]

Originally posted by: leeland
In the event I was able to get a router/switch that did support JUMBO frames how much of an impact would that have as far as improving the speed

vs.

A standard router/switch that didn't support JUMBO frames?

To your NAS, probably not much. As was already pointed out, SOHO NAS devices are notoriously slow, bottlenecked not by network bandwidth, but rather by disk access speeds.

In my opinion, don't waste the money.

 

[Madwand1]

Originally posted by: leeland
I will change things back to the default and see what kind of speeds I get

The DGL-4100 would simply drop jumbo frames, so you couldn't be transmitting jumbo frames, so changing the settings now should have no real effect as jumbo frames are already disabled somehow.

Originally posted by: leeland
In the event I was able to get a router/switch that did support JUMBO frames how much of an impact would that have as far as improving the speed

vs.

A standard router/switch that didn't support JUMBO frames?

Interestingly, SNB's NAS charts have data on this if you look closely. (I only consider the "1 G" file size in their tests, as the others just have more impact from caching and don't represent sustained transfer speeds as well.)

RAID 1 write with 4K jumbo frames: 16.8 MB/s
RAID 1 write with no jumbo frames: 13.7 MB/s

RAID 1 read with 4K jumbo frames: 18.8 MB/s
RAID 1 read with no jumbo frames: 15.8 MB/s

So jumbo frames give around 20% improvement here. A jumbo-capable gigabit switch such as a D-Link DGS-2205 can be pretty inexpensive if you shop well. Whether or not it's worth that and the potential hassles with jumbo frames for ~3 MB/s improvement is your call. (Most people here would say no.)

Note also that there's some guesswork here -- SNB's figures are interesting as they also match your ~13 MB/s write, but there's no real guarantee that you'd get the same improvement with jumbo frames as your overall system is different.

 

[Madwand1]

BTW, there's a fairly simple way to test jumbo frame performance without a jumbo-capable switch -- just use a direct cable connection (most gigabit is auto-crossover, so a standard cable works fine). You might have to manually set the IPs or adjust for them in this case as your router's DHCP wouldn't be available.

 

[leeland]

Thanks again for the posts and the tips...it has brought up some excellent points and I will run some tests to see if I can get any improvement.

I would be happy if I could get it back up to the sustained speeds I had it earlier of 13 to 14 mb/sec since I started playing around with it and changing settings it has gotten worse ( go figured lol )

When I get some free time, maybe tonight I will try to change things back to the way they were and see if I can get any improvement