Office Ethernet Speeds

[CrippledROBOT]

Hello everyone! A quick question about our office internet situation. We have installed CAT6e lines throughout the office. We also have our server switches rated for gigabit transfer (Rosewill RNSW-1101). Most of the motherboards that we have in our towers have gigabit LAN ports. However when I use a software such as LAN Speed Test I will get speeds around 400-600mbps rather than a full gigabit speed. I assume that this means we have a bottleneck somewhere? If so, what might it be?

[Electronics Wizardy]

netwrk diagram would help, its probably limited by load already on the network.

 

what speed do you get with 2 computers connected to a switch and nothing else?

[CrippledROBOT]

Just now, Electronics Wizardy said:

netwrk diagram would help, its probably limited by load already on the network.

 

what speed do you get with 2 computers connected to a switch and nothing else?

I will definitely try that. Now that you mention that, this would make sense. I work at a small accounting firm and we are in the throws of tax season so I will have to test this when we have less network use across the network and have a chance to test with two computers connected to a switch and nothing else. Assuming that the speeds are exactly the same in this test, what else would you suggest?  

[Electronics Wizardy]

Just now, CrippledROBOT said:

I will definitely try that. Now that you mention that, this would make sense. I work at a small accounting firm and we are in the throws of tax season so I will have to test this when we have less network use across the network and have a chance to test with two computers connected to a switch and nothing else. Assuming that the speeds are exactly the same in this test, what else would you suggest?  

what exact switch googling that model brings up nothing, and im guessing thst may be a issue.

[CrippledROBOT]

Just now, Electronics Wizardy said:

what exact switch googling that model brings up nothing, and im guessing thst may be a issue.

https://www.newegg.com/Product/Product.aspx?Item=N82E16833166071

[Electronics Wizardy]

5 minutes ago, CrippledROBOT said:

https://www.newegg.com/Product/Product.aspx?Item=N82E16833166071

looks ok, but next time id get a managed switch from a name brand, makes the network easier to manage.

 

My guess is you have lots of systems using it and lots of load going over a single link, and thats your problem.

[CrippledROBOT]

1 minute ago, Electronics Wizardy said:

looks ok, but next time id get a managed switch from a name brand, makes the network easier to manage.

 

My guess is you have lots of systems using it and lots of load going over a single link, and thats your problem.

I personally wouldn't have gone for that unit, but that's what my boss and his close friend (who also happens to be our part-time IT support) had gone with. I came up with a similar conclusion, good to hear that others agree! Thank you very much for the help!

[Blebekblebek]

Switch is a switch, every port is independent. So unless that server test is getting hit with another machine then the speed should be a gigabit or at least close to it.

 

the issue here isn't only about the unmanaged or managed switch.

 

do iperf test properly, there's a lot limitation if you do the conventional test with HDD with windows machine, since windows it self have QoS to shape your traffic.

 

The idea to get "managed" switch just to achieve gigabit speed is laughable.

 

[CrippledROBOT]

1 minute ago, Blebekblebek said:

Switch is a switch, every port is independent. So unless that server test is getting hit with another machine then the speed should be a gigabit or at least close to it.

 

the issue here isn't only about the unmanaged or managed switch.

 

do iperf test properly, there's a lot limitation if you do the conventional test with HDD with windows machine, since windows it self have QoS to shape your traffic.

 

The idea to get "managed" switch just to achieve gigabit speed is laughable.

 

What do you recommend that I test/what is your recommended testing procedure?

[Blebekblebek]

2 minutes ago, CrippledROBOT said:

What do you recommend that I test/what is your recommended testing procedure?

https://www.sd-wan-experts.com/blog/iperf-bandwidth-testing/

 

[CrippledROBOT]

9 minutes ago, Blebekblebek said:

https://www.sd-wan-experts.com/blog/iperf-bandwidth-testing/

 

Thanks! I'll give that a try! 

[Mikensan]

Every port is not independent on cheap switches, usually they share a controller. Maybe 4 ports to a controller or worse. This ends up with switches having a total real throughput of a gigabit/s on extremely cheap ones. 24 ports for $70 brand new is really-really cheap. It is a little sketchy in the specifications also... Some of the 24 port cisco switches (ie 3550) don't hit 48gbps. Granted this is just a L2 and should have higher speeds, it just doesn't add up.

 

Another possible issue is if the cable you ordered actually says "6e" you might have some odd-ball cables, as @dalekphalm pointed out to me in another post.

 

All that aside, I'd also suggest using iperf to test between computer/server. In the event you have a bad disk on either end that could be your bottleneck, iperf should use RAM in its testing (if I recall correctly).

[CrippledROBOT]

3 hours ago, Mikensan said:

Every port is not independent on cheap switches, usually they share a controller. Maybe 4 ports to a controller or worse. This ends up with switches having a total real throughput of a gigabit/s on extremely cheap ones. 24 ports for $70 brand new is really-really cheap. It is a little sketchy in the specifications also... Some of the 24 port cisco switches (ie 3550) don't hit 48gbps. Granted this is just a L2 and should have higher speeds, it just doesn't add up.

 

Another possible issue is if the cable you ordered actually says "6e" you might have some odd-ball cables, as @dalekphalm pointed out to me in another post.

 

All that aside, I'd also suggest using iperf to test between computer/server. In the event you have a bad disk on either end that could be your bottleneck, iperf should use RAM in its testing (if I recall correctly).

Wow thank you! You have all been incredibly helpful! I admit that this is not my particular area of IT expertise. I'll give that a try tomorrow!