NIC card is confusing me...

C

ct_hamster

New Member
Hello,

Does anyone know of/knows a website that explains the difference in NIC card speeds? I have three Server computers (one Windows 2k3 the other two centos 5.3) and they each have different NIC cards. The odd thing is how they perform.

Main computer (lowest results in local network test)
OS: Windows 2K3

4 GB RAM
C Local Drive 14.6 GB (9.1 GB free)
D Local Drive 683 GB (540 GB free)

INTEL(R) PRO/1000 EB Network Connection with I/O Acceleration
Speed: 1 GB/s

CAT 5 network cable

------------------------------

2nd computer (Better results in local network test than main computer)
OS: Centos 5.3

4 GB RAM
Local Drive 17 GB (8.8 GB free)

Broadcom Corporation Netxtreme BCM5720 Gigabit Ethernet PCI Express
Speed 1Gb/s

CAT 5 network cable

------------------------------

3rd computer (Over triple speed of 2nd computer in local network test)
OS: Centos 5.3

1.5 GB RAM
Local Drive 373.7 GB (465.66 GB free)
Intel Corporation 82557/8/9/0/1 Ethernet Pro 100
Speed: 100 MB/s

CAT 5 network cable

------------------------------


Program used: http://download.cnet.com/LAN-Speed-Test/3000-2085_4-10908738.html
 
First of all, the first two servers you list have Gigabit NICs in them. Their peak throughput (physical limit of the card, not real-world) would thus be 1Gb/s. The third server you list only has a 100Mb NIC in it. It will have a physical limit of 100 Megabits a second.

Second, you list all your cabling as CAT5. Unless this is at least CAT5e class cabling, you will notice very odd behavior on your gigabit backbone on runs over 50 feet - provided you have high quality cable within shielded conduit. Standard CAT5 ethernet cable does not have the proper shielding or frequency bandwidth to handle Gigabit speeds.

Third, please describe your whole network topography between these units. Multiple routers, switches and hubs can severely limit your throughput depending on what you have configured.

Also, I am not familiar with this program, but it may also take into consideration the speed of your hard drives. If any of those systems are particularly bogged down with other tasks, you could see vastly differing results. The best way to test network speeds is between two known entities.
 
It appears that the cords used are CAT5e not CAT5. My boss wasn't here for me to ask and I didn't know the difference. sorry.

All three computers are connected to a Linksys SR2016 Gigabit Switch (16 port) and the other computers in the network are connected to a second Linksys SR2016 Switch which is linked to the first one. The first Linksys SR2016 Switch Is connected to the router.

Router
Linksys SR2016 Switch
Servers
Linksys SR2016 Switch
Wall
Networked computers

Hope that makes sense...

We were going to switch the CAT5e Cords for CAT6 cords. Would that be any more effective with network traffic?
 
That makes sense. How were you conducting the test? From server to workstation? Server to server? Loopback?

CAT6 can make a marked improvement especially if you have your data lines next to other noisy RF sources like power lines and lighting fixtures. CAT6 is certified for Gigabit use and highly recommended.
 
Have you tested from the same workstation to all three servers multiple times when there is no other network activity?
 
The computer I am using has a 100 MB/s connection. I have tested multiple times with as little network activity as possible and still received approximately the same results. I also ran the test from a computer with a 1 GB/s connection and still got similar results.
 
OK. What is the average throughput you are seeing for each server from that workstation? Your original post is not specific.
 
Results based on program mentioned before

Main Computer

Writing
Packet Length:104,857,600
Time to complete (in seconds):34.9690000
Bytes per Second:2,998,587
Bits per Second:23,988,696
Mbps:22.8774033

Reading
Packet Length:104,857,600
Time to complete (in seconds):9.2180000
Bytes per Second:11,375,309
Bits per Second:91,002,472
Mbps:86.7867203

-----------------------

2nd Computer

Writing
Packet Length:104,857,600
Time to complete (in seconds):35.9070000
Bytes per Second:2,920,255
Bits per Second:23,362,040
Mbps:22.2797775

Reading
Packet Length:104,857,600
Time to complete (in seconds):9.2970000
Bytes per Second:11,278,649
Bits per Second:90,229,192
Mbps:86.0492630

-----------------------

3rd Computer (Fastest Results)

Writing
Packet Length:104,857,600
Time to complete (in seconds):14.0630000
Bytes per Second:7,456,275
Bits per Second:59,650,200
Mbps:56.8868637

Reading
Packet Length:104,857,600
Time to complete (in seconds):11.3280000
Bytes per Second:9,256,497
Bits per Second:74,051,976
Mbps:70.6214676

-----------------------

The more I run the tests the more rnadom they seem to get...
 
Those look rather consistent from my perspective. I am looking at the Mbps field. Are you comparing the time to complete? Note that computers 1 and 2 are very similar and that 3 has the longest read time and the shortest write time.

Double check that all the systems that are supposed to be gigabit (servers and workstations) are actually connecting to the switch at gigabit speeds. The switch should give you some notification of this.

I am not familiar with this tool that you are using and how it operates. You may want to look into something like PassMark's network performance tools. Also, have a read through this article for some good information. It is a bit old, but has some relevant info yet.
 
You must log in or register to reply here.
Back
Top