Is there a hard drive independend way to test LAN bandwidth?

[Staples]

I used to have a 100mbps network until a week ago. I was getting about 80mbps, 80% of what I should have been getting.

I now have a cheap gigabit Netgear switch, cat5e (350mhz) cables. In one computer, I am using the integrated nforce 3 LAN and in the other computer I bought a full duplex gigabit PCI adapter. All equipment is saying that the connections have been detected at gigabit connections and not 100mb.

Surprisingly, my bandwidth between these two computers is only 110mbps. That is less than 50% more than what I was getting before. Both hard drives should be much faster at reading/writing than 13MB per second but a hard drive bottleneck is the only limitation that I can think of.

Is there some type of test where I can send data from one computer to the other without being limited to the hard drive speeds? Like maybe one computer sends 200mb of data from memory to the other computer?s memory. I am sure that would max out the gigabit bandwidth and determine if the hardware was faulty or not.

 

[JackMDS]

This wouls shade some light on Giga.

Peer to Peer Giga Networks

Setting a Home/SOHO Giga network.

Measuring Transfer, http://www.ezlan.net/faq.html#transfer

:sun:

 

[Staples]

Originally posted by: JackMDS
This wouls shade some light on Giga.

Peer to Peer Giga Networks

Setting a Home/SOHO Giga network.

Measuring Transfer, http://www.ezlan.net/faq.html#transfer

:sun:

That sheds some light on the subject. I never did have expectations of real gigabit bandwidth because no hard drive is that fast and the wires are only rated for 350Mbps. I was however expecting for the hard drives to be the bottleneck (so 25MB per second or so). Guess I had too ambitious of expectations. I do enjoy my 35% boost but I was hoping for more.

Oh well, as the article says, when OSes evolve, I should be able to get more speed out of the network. And unfortunately, my switch does not support jumbo frames. But I don't think I'd enable it even if it did because every device on the network has to support it which it does not.

Do you think I would get much of a boost if I did have jumbo frames compatible hardware with my current max being 110Mbps?

 

[Madwand1]

I'm getting around 30 MB/s file transfers using built-in NICs through a consumer switch without jumbo frames, IDE to IDE. When the sending side is cached, I get 50-60 MB/s.

These particular measurements were for a roughly 1.25 GB directory with files averaging around 8 MB each.

I think you should be able to do better.

SiSoft Sandra gives simple network benchmarking among a number of other potentially useful tools. I certainly cannot certify this as a reliable (or even very flexible) tool, however, it's easily accessible and usable. Its reported measurements are not far off what I get, but big grains of salt are needed for good taste. In this case, Sandra says I get 62 MB/s transfer speed from the source computer to the destination, which happens to roughly match what I see in the file transfer performance.

I'm not going to hazard a guess as to what the problems are in your case, but raw gigabit throughput and modern HD's can give accessible significant transfer performance, though not at full gigabit speed of course, as my case shows.

Microsoft has a couple of guides for improving network and file server performance -- do a search for Microsoft file server performance or variations, and you should be able to find them.

 

[Staples]

Well this is interesting. I downloaded Sandra and tried it on both computers.

I assume that the number you get is from your computer's ability to send down the pipe to the host computer. The test package that is sent is under 1kb.

Anyway, one computer claims 27Mbps and the other brings back 23-24Mbps. Either way, both are faster than I got transfering one big file (which is the way I tested this in the first place and not several small files like the poster above). I will try some of the optimization tips and see if I can get it any faster now that Sandra is giving me some hope. Both computers are very fast systems so it isn't totally clear to me why IDE to IDE is 13Mbps.