Sign in with
Sign up | Sign in
Your question

Expected data transfer speeds on 10/100 network

Last response: in Networking
Share
April 9, 2007 1:44:04 PM

Hi!
I'm trying to figure out what data transfer speeds are typical on a 10/100 network.

My home network is currently setup as:
Modem to Linksys BEFWS114 router (DHCP enabled) to Dell PowerConnect 2216 (unmanaged switch)

Cat5e cables used all thru'.
I have one PC (lets call it PC1) (full duplex) connected to the BEFWS114 and another (lets call it PC2) on the 2216 switch.

Last weekend, I was trying to transfer a data file (5GB file) using a windows explorer drag/drop from the PC1 to a share on PC2 - both PCs are running Win XP Pro SP2. The transfer was slower than what I was expecting - it was close to 12Mb/s. With a 10/100 network, I was expecting at least 40-50Mb/s.

Now, I unplugged the PCs from the network and connected them directly using a crossover cable. Assigned IPs (same subnet) and tried to transfer the same file over - it was still stuck between 12-15Mb/s.

Could someone help identify what I'm doing wrong?
Both the PCs are using the integrated NICs from the mobo (ASUS P5W-DH in PC1-10/100/1000 & MSI NEO2PLS in PC2 10/100)

Any help is appreciated!
Thanks
April 10, 2007 5:08:34 AM

Don't confuse 'b' and 'B'.

B = Byte = 8 bits = 8 b.

So 12 MB/s = 12 * 8 Mb/s = 96 Mb/s which would be very nice 100 Mb/s performance. If you didn't mean byte, then using B is just confusing.

Edit: Misread OP regarding second NIC - disregard direct connection suggestion here.
April 10, 2007 11:30:40 AM

Sorry about that - I have edited my post to indicate Mb/s (instead of MB/s).

One other post I read recommended using FTP to connect between the PCs - I'm not sure about how to do that - how can I do that? (and can it really improve performance)?

Is there any freeware out there that will monitor n/w traffic for dropped packets or other such issues?

Thanks
Related resources
April 10, 2007 3:51:57 PM

How long did it actually take to transfer your 5GB file?
April 12, 2007 12:25:05 PM

I actually stopped the transfer after about 20 minutes, so I really don't know how long it would have taken...
Do you know of any tools (software) that can monitor network performance? (dropped packets, retransmits etc?)
April 12, 2007 2:15:18 PM

Quote:
I actually stopped the transfer after about 20 minutes, so I really don't know how long it would have taken...
Do you know of any tools (software) that can monitor network performance? (dropped packets, retransmits etc?)
Just some numbers (just 'cause I have to...) :wink:

Assuming the file was 25% complete, that is about 8Mbps. Pretty poor.
Assuming the file was 99% complete, that is about 30Mbps. Even at that, something seems amiss, if you were actually measuring network performance.

With a file that large, you are also measuring sustained transfer rate of the hard drives of both machines, plus the overall PC performance on both ends. Still, unless your drives and PCs are real dogs, it should have been higher.

The only network performance tools I am familiar with are very expensive commercial IT tool sets. There are probably consumer level or hobbyist level network monitors, but I don't know where to point you.
April 12, 2007 6:25:18 PM

IceBlue, thank you for your response. windows progress bar indicated the transfer to be close to 25% - that's why I ended up canceling it out.
Both the PCs have fairly decent configurations
PC1 is running a E6600 C2D on a ASUS P5WDH board with 4GB DDR2-800 CRUCIAL Ballistix RAM - I have a couple of 150GB Raptors on Raid1 config on the Intel bridge (not the SIL)
(I know RAID1 is going to be a little bit slow, but should not be slow on reads)
PC2 is running a 2.8GHz P4 Prescott on a MSI Neo2PLS with 2GB Crucial RAM - Source drive is a single Seagate 320GB 7200.9

Besides, even with a direct connection using a crossover cable, the speed didn't improve... There's gotta be something wrong that I am doing... something with the way the PCs are setup - I just have no clue what! Grrrrr!

Thanks anyway. If you think of anything else, please let me know.
April 14, 2007 7:47:57 PM

You can use iperf for network performance tests. I suggest version 1.7.

http://dast.nlanr.net/Projects/Iperf/

E.g.

server: iperf -s
client: iperf -c server -l 64k -t 12 -i 3 -r

This will give you a sense of the "raw" TCP/IP performance. This will factor out the hard drives, the file system, caching, and also the composite load of the drive subsystem together with the networking. It will also factor out the file transfer utility.

FTP isn't hard to set up, and comes included with XP Pro. However, XP Pro itself has some issues with native file transfer performance, and I believe this is not entirely bypassed when you use its native FTP implementation. For this, I'd suggest going to a third-party tool. I haven't really investigated these, but have gotten very good performance from FileZilla Server, at the cost of high CPU utilization at times. The FileZilla client is likely to be a decent choice (not the best perhaps, but convenient and at least passable in terms of performance). I suggest reading the docs for setup help. (To be honest, I winged it, but it wasn't intuitive, and I didn't select the best configuration; just something good enough to get by for testing.)

You also have to remember to let FileZilla/ftp go through the Windows firewall if you're using it. This brings to mind that for diagnosis of performance issues, you should try disabling the firewall and any virus scanners, ensure that you don't have any wireless perchance (which I don't think you do, but am guessing wildly based on the poor performance), etc.

The other wild guess I have is that one of the drives may be degraded to PIO mode. This would appear as high CPU utilization and poor disk performance. Tools like HD Tach can give simple and easy-to-use read performance measurements. Getting more thorough (including write performance) would be better, but starting simpler is OK as far as it goes.

So above, you have means to:

1. Isolate network performance
2. Isolate file transfer protocol performance
3. Isolate drive performance

This should give you some info.

Personally, I'm a gigabitbigot, so would jump ahead and install a PCI GbE NIC in the 10/100 computer, and add a GbE switch, but, but.. (1) There is some problem in your case that might kill that idea. (2) There are some limitations in the performance of XP Pro (which is addressed in 2003, XP-64, Vista), and so some chance of additional frustration there -- this is the problem that you could potentially avoid using FTP. There are also some settings that can help, but nothing that would bring down 100 Mb/s performance to 6 Mb/s levels as far as I know, which is why I suggest the more detailed analytical approach first.
!