• 09/03/2015
    1:00 PM
  • Rating: 
    0 votes
    Vote up!
    Vote down!

Network Troubleshooting: Latency Vs. Retransmission

In this video, learn how changing the location of network test tools can help identify problems with latency and retransmission.

When troubleshooting network performance problems, most analysts find themselves chasing one of two issues: latency or retransmissions. Both scenarios result in performance degradation, but have very different root causes and solutions.

I've been involved in many troubleshooting exercises where I get a totally different perspective by changing my test point. In this video, I'll show you how the same problem can look like a latency problem or retransmission-related issue by simply changing your location in the network.

Here are some additional tips when trying to identify if an issue is latency or retransmission related:

  • Try to monitor from the sender's perspective. If the sender is not physically close, then make yourself the sender by uploading a file or running iperf.
  • Pay attention to your protocols. This information presented here is helpful when dealing with TCP base protocols. UDP is a totally different animal.
  • Try to leverage operating system commands like netstat –s to identify retransmissions.
  • Understand what your tools are reporting. For example Wireshark might note a retransmission, spurious retransmission, fast retransmission, or other notes.
  • Look for TCP-specific hints like Selective Right Edge (SRE) or Selective Left Edge (SLE) to possibly indicate packet loss.

There is no guaranteed tip or trick when troubleshooting packet loss or latency issues. But awareness of tool placement will help you when troubleshooting.

>> See Tony discuss this issue further and answer audience questions at Love My Tool.



Thanks for this helpful video Tony. You mention that UDP is a different animal when it comes to identifying latency or retransmission; can you offer any basic tips on that? 


What was that?  Can you repeat the question? LOL...  I couldn;t resist some retransmission/geek humor ;-)

Here's some background information.  

Since TCP has a built Sequence, ACKnowledgement numbers, and other features retransmissions are easier to spot and deal with.

UDP has no such built in mechanisms, so the application has to figure it out.

For UDP packet loss and out of sequence analysis I go to the IP layer where there is a sequence number (identifier) to get an idea of how things are going.



LOL! Thanks Tony!


It explains most of the point when we say TCP is connection-oriented and UDP is a connectionless. I also dont want to miss UDP is still preffered over TCP for applications that need fast transmission, such as games.


Excellent points.  Thanks for taking the time to share.




Fair enough statements you have made here.

This is precisely why UDP is used for High-Speed Gaming Applications as well as applications like HFT where you need Rapid Bursts of Data without worrying about Connectivity Infrastructure perse(its taken for Granted).

Good basic Primer you have shared with us here at NetworkComputing!



good feeback, thanks Ashu001,

thank you

great video, thanks!  Love this level of simplicity + detail.

Re: thank you

And thank you kbenner33301,for taking the time to share some feedback.

Much appreciated