One-Way Ping Jitter Test

0

Would a public website like I describe here have any value? Does something similar already exist? Right now, I'm using tracert to debug a network problem in which there is a delay at times between two internet backbone routers, but would further like to know which direction is causing the delay. I even see an increase at times in packet loss and want to know which direction is losing the packet. By the way, I only have my one computer, so using iperf isn't an option.

So, I envision a public server (just a website that I open in any browser) sending me one datagram every 5.00 seconds (i.e., between 4.99 and 5.01). My computer immediately responds, by sending the datagram for a TCP ACK packet, so the server's next datagram can thereby include it's measured "ping" round-trip time. (To be clear, in case my TCP response is that of an error, the datagram re-send from the server should be sent after waiting the full 5.00 seconds; the key is that I want a reliable datagram to arrive every 5.00 seconds, so it's fine even if the server simply ignores my TCP responses.)

If my browser then shows me the two separate jitters (for datagram arrival times and "ping" round-trip times), I can easily see which network direction is causing the jitter. It's important to realize that my network problem comes and goes at times, so I would simply load this website during short delays (confirmed because the shown "ping" round-trip time will be at its minimum) and watch the reported times change when the long delays start. If delays were always high, this website would not help.

bobuhito

Posted 2017-10-25T08:10:05.013

Reputation: 131

I don't think this approach is going to work, as there is no guarantee that routing is symmetric. See my answer Why doesn't "ping -j" work? for a more detailed explanation, in particular "What is a reverse route?"

– DavidPostill – 2017-10-25T15:49:46.460

Even if routing is not symmetric, I would still learn which direction is causing the jitter. And, if tracert were run from the server to my computer (in addition to tracert being run from my computer to the server), I would know the separate routes for both directions and find the exact slow link for the bad direction (by looking at the tracert times). So, I'd say that this approach still works. – bobuhito – 2017-10-26T00:27:54.143

No answers