Latency of 40ms drops total server bandwidth

Started by jkiac


Latency of 40ms drops total server bandwidth   13 February 2014, 17:33

i have a virtual machine which acts as a gateway between internal servers and a cloud service. We are testing latency and packet loss to the cloud service.

when setting up latency on either internal server facing NIC (NIC1)or cloud service facing NIC (NIC2), the bandwidth is dropped to a value not acceptable. A 40ms outgoing latency on NIC1 or NIC2 OR 20ms both, drops the 100 Mbps bandwidth to just 10 Mbps. When testing with other WAN emulator, this does not happen.

Any insights?
SoftPerfect Support forum - Andrew avatar image

Re: Latency of 40ms drops total server bandwidth   13 February 2014, 17:39

It sounds about right. Please consider this article. The author found the following correlations (round-trip latency/TCP throughput)

  • 0ms 93.5 Mbps
  • 30ms 16.2 Mbps
  • 60ms 8.07 Mbps
  • 90ms 5.32 Mbps

So 40ms round-trip-latency or 20ms + 20ms slows a TCP connection to about 13 Mbps.

The article How to Calculate TCP throughput for long distance WAN links provides a few formulas that lead to the same numbers.

Perhaps the other emulator runs on a different system with a different TCP window size?

Hope this helps.

Reply to this topic

Sometimes you can get the answer faster if you try the forum search and/or have a look at the software user manual to see if your question has already been answered.

Our forum rules are simple:

  • Be polite.
  • Do not spam.
  • If possible, check your spelling and grammar.




A brief and informative title for your message, approximately 4–8 words:


Spam prevention: please enter the following code in the input field below.

 ********  **         *******   ********  ******** 
    **     **        **     **  **           **    
    **     **        **         **           **    
    **     **        ********   ******       **    
    **     **        **     **  **           **    
    **     **        **     **  **           **    
    **     ********   *******   ********     **