Howto assess Network performance?

[Peroxyde]

Hi,

I am joining a project where the HyperV infrastructure experiences some hard to find slow performance issues. We would like to breakdown the investigation into networking and disk subsystems. As this is the networking section of the forum, can you please advise a testing or benchmarking tools to assess or compare network performance? For example:

- Between Machine1 and Machine2 outside of HyperV
- Between Machine1 and VM1

Hoping that this could help us to narrow down the investigation to some identifiable causes.

Thanks in advance for any advice.

 

[spidey07]

Normally you'd use MRTG or some other SNMP tool to measure bandwidth utilization on the network ports. Even the command line of the switches can tell you if there are congestion/drops.

First place to look though is cabling and duplex mismatches. That's the #1 cause of performance trouble. It's rare that you can overload a properly designed data center LAN. But if the cabling isn't up to snuff, then all kinds of weird performance problems happen.

 

[Cooky]

What spidey mentioned is general network performance on a Layer 1 - 3 perspective.
Usually even if you tell the application & server folks everything looks ok on your side isn't enough, and it sounds like that's where you are, or you double duty as the server admin.

On a high level, what you need is perform packet captures from point A to point B, and use tools such as Netscout's InfiniStream to analyze data, and see where the delay occurs.
What I usually do is meet them half way, "my network looks fine, but here's the packet capture for you to figure out why your app isn't working right".

 

[spidey07]

Lol. Then the blank state when they look at the trace.

Ping pong. Ping pong. Why is the network so slow?