Bad Network Response on IBM X Series 345

[stevehynes]

We have a new IBM X345 series server and the network performance of it is terrible. We have a static IP address and is using a teamed NIC.
Any ideas as to what could be causing this

Steve

 

[catorze]

Could be any number of things. What is the server connected to? Is everything configured to Auto-negotiate? What else is going on in the network? What are you using to quantify network performance? What are you teamed with? Are you configured for load balancing or fault tolerance? The x345 has an integrated 10/100/1000 Ethernet controller and should provide exceptional performance, provided your network can support the speeds.

 

[stevehynes]

Cheers.
The server and all the PC's are connected into a Cisco switch. The port settings are as follows,
Duplex = auto, Speed = auto, Port fast = enabled.
Is the fact that the settings are auto configured affecting the network speed. Should I change the duplex to full and the speed to 100MB where needed?

 

[catorze]

No, auto is the correct setting, provided the entire network (or at least everything attached to the same port) is set to auto-negotiate. If you are forcing a speed or duplex mode on one of the ports (or any system is configured for a specific speed or duplex setting), then that port should be hard set to match. Auto-negotiation usually will only successfully negotiate if both ends are configured to auto-negotiate. If it cannot negotiate, then it will default to the lowest speed/half-duplex. This usually creates a problem if negotiation fails, because the switch is talking at 10Mbps/Half and the system is at 100Mbps/Full. Consequently, they will not communicate.

 

[xmsre]

Take a sniff. If you see a lot of runts, the issue is most likey one with adaptive transmit threshold. On the properties of the NIC, advanced page, set adaptive transmit threshold to a value of 200.

http://www-1.ibm.com/support/docvie...g1MIGR-43673&loc=en_US&cs=utf-8&cc=us&lang=en

For an explanation of advanced settings, see:

http://www-1.ibm.com/support/docvie...1MIGR-4GYSMK&loc=en_US&cs=utf-8&cc=us&lang=en

John
MOSMWNMTK

 

[dhauser]

I would suggest that you check to see if you have the latest BIOS update from IBM. These servers are know to have issues with those NICs if the BIOS levels are not up to date.

If that does not fix it try change out the network cable. Remember the simple answer is most offten the right one.

Dan

 

[helpless1976]

I am having the same network performance problem. I have 1 Server running on Windows NT with 10 clients heavily accessing it. I am using a 3com Switch with the following settings on the Server's NIC:

Adaptive Interframe spacing = 1 (default)
Adaptive performance tuning = Max adapter bandwidth
Adaptive Technology = on
Adaptive Transmit Threshold = 16 (default)
coalesce buffers = 32
Link Speed = auto detect
PCI Bus efficiency = disabled
Receive buffers = 1024
respond to flow control = disabled
transmit control blocks = 64

Though performance had improved since I maximized the Receive buffers, coalesce buffers and transmit control blocks, I still have occassional failures on the LAN.

Should I set Adaptive Transmit Threshold to 200 and enable the respond to flow control? or I got the settings all wrong?

thanks

 

[dhauser]

To Helpless1976

You are running NT 4 server on a server that does not support NT 4.

Think that may be part of your problem.

http://serverproven.raleigh.ibm.com/compat/nos/microsoft.html

Dan

 

[helpless1976]

Dan,

I am using a Primergy, Fujitsu Siemens Server which supports NT 4 Server.

It was running well when the load on the network was low. After increasing the load, I had to replace the hub with a switch because i am getting excessive collisions. But, I could not get the right settings on the Server NIC.

Do you have any idea how to improve the performance?

Thanks,
Elea