Bandwidth Tests

netstat

Hi guys, hope you all k!

I am searching for a free reliable bandwidth tester. I have used the below so far and both gave me different readings.

-Mikrotik
-TTCP

TTCP is one which i knew was reliable although i would like to know what you guys use.

These are the settings i used for TTCP

C:\Bandwidth Speed tests\ttcp>pcattcp.exe -r 100.100.100.1

C:\Bandwidth Speed tests\ttcp>pcattcp.exe -t -l 100000 -n 1488 100.100.100.2

To make sure that the tester is working correctly (to test the tester) i started out with 2 laptops and a crossed cable. i never managed to get very close to 100Mbit (got 85Mbit). I played around with duplex mode and the frame buffs of TTCP but 85 was the max i obtained!
I can blame the laptops although i have tried others and i got the same result more or less. The cable is 1/2 a meter long and tested so i am out of ideas....

if u know of any good testers let me know.

Cheers!

Mishra

You are never going to get 100 Mb out of a connection. They sell it as that high but you never get it.


Honestly when I do bandwidth tests I do it with the particular item that I need to test... What I'm trying to say is if I need to know how fast files are going to transfer over to my other windows machine then I just do a test transfer and see what kinds of speeds I'm getting and how long it takes.

Bandwidth testers don't take into fact the overhead that windows file transfer adds into the time or whatever you are doing over the network. Usually they try to streamline the process to get as fast as you can transfer at which isn't real life measurements.

netstat

hey thanks for your reply,

I was not really expecting 100Mb but 15% loss was a little too much to believe. In fact i repeated the test with fresh installed PCs, The others had some group policy applied since they were part of the domain - but were connect back to back for the test as previously mentioned - no firewalls or running programs also. I manged to get it to 92Mbit, which is more like it. So weirdly enough, having Pc with an applied GP/Part of a domain - affected my bandwidth!

JDMurray

There is a difference between the total bandwidth of a connection and just the data throughput. If your bandwidth tester is only measuring data throughput, it is only considering the size of packet payloads and not the entire packet itself. True connection bandwidth testing is performed using streaming and not packetized data. This is why Internet speed tests are only good for discovering current data transmission rates and not the absolute maximum bandwidth of a connection. You need to find out exactly what your speed tester is measuring.

netstat

yes yes i know exactly what you mean. I am actually more at a frame level here. No IP / tcp headers cos the result would vary in that case - It is not a realistic bandwdth result i know! But just wanted to see if a cable or fibre optic is rated for that speed, what other losses are we to expect? Assuming terminattion, and all cable losses are set to nill? Ethernet packet is approx 1500 bytes , i am testing in the 1488 bytes range, this should be close enough..

Cheers!

dtlokee

15% is about right. There are many thins at play here that will reduce the total speed. First there is additional framing overhead adn delays between frames that will prevent you from hitting 100%, that will drop it to about 95%, after that you need to start looking at the internal architecture of a PC and how a NIC attempts to saturate a link. The concept is there must be something to send in the output buffers at all times which is acheived by using shared memory and DMA, but what if the bus is currently busy? what is the maximum performance of the PCI bus? What latency is introduced by moving data from memory through the chipset to the NIC?

Try a similar experiment on a few similar computers with gigabit interfaces and see what the result is. You won't hit 850mb/s (85% of 1000mb/s) due to limitations of the hardware.

netstat

agreed! I had problems with some hardware at a point too (before i got the 92Mbit) . For example, if both nics are auto duplex i get 3Mbit! with one Auto and the other 100Mibt fixed i got 80Mbit! - drives me crazy! With gbit i got nasty results too, but here kicks in you point that the hardware maybe wasn't in its ideal state for such a test - Fully patched up windows XP, with added AV ....