[tforum] Fwd: [sciencedmz] impact of a small amount of packet loss on throughput

Joe Breen joe.breen at utah.edu
Fri May 24 12:13:49 MDT 2013


For those troubleshooting performance or wondering how dirty networks
impact traffic, the below message link and associated info might be of
interest.

To give a feel in terms of scale, 13-15ms round trip time latency allows
you across the state of UT (north/south) on UEN's backbone.  25-30ms of
round trip latency allows you to Kansas or Houston.  50-60ms of latency
allows you to travel to/from Atlanta or New York from the UofU.  159ms of
round trip latency allows your packets to visit London and back.  160-170ms
of round trip latency will allow your packets to visit Beijing, China.
229ms of round trip latency will allow your packets to travel to Bangalore,
India and back.

  --Joe

---------- Forwarded message ----------
From: Brian Tierney <bltierney at es.net>
Date: Fri, May 24, 2013 at 11:37 AM
Subject: [sciencedmz] impact of a small amount of packet loss on throughput
To: Science DMZ List <sciencedmz at es.net>



Folks on this list might find this interesting:

>
http://fasterdata.es.net/performance-testing/perfsonar/troubleshooting/packet-loss/

No surprises here, but the plot shows pretty dramatically how if you are
trying to
move data more than 10ms away using TCP, you need a clean path!




_______________________________________________
This is the sciencedmz mailing list.
To subscribe or unsubscribe to this list, go to
https://listserv.es.net/mailman/listinfo/sciencedmz
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.uen.org/pipermail/tforum/attachments/20130524/31bb481f/attachment-0001.html 


More information about the tforum mailing list