Question

Is periodic high latency typical or a technical failure?

  • 7 January 2024
  • 6 replies
  • 181 views

Badge

Our TMHI Arcadyn gateway is connected to a pfSense router and feeds our home network via a wired network. I have the on-board wifi disabled and don’t use it. The tower is 3,200 feet away. Signal strength is ‘Very good.’

Normal latency averages about 40ms with a standard deviation about 10ms.

Ever few weeks, however, latency becomes unstable for days and averages around 80ms with a typical standard deviation average of 60 to 80ms. Latency can spike as high as 150+ms. Sometimes there’s a build up of instability at a lower level until a bigger episode. I’ve attached a graph of the last month, which is typical.

This issue has occurred on more or less random intervals varying between one to two months. While bps speed remains more or less the same, the high latency periods decimate responsiveness for the days it takes to resolve itself. And it does resolve itself: even if I do nothing to the network the issue appears and disappears on its own after a random period of time from hours to days.

Multiple calls to support were fruitless. They don’t seem know what I’m talking about and seemingly can’t understand what latency means. One tech even tried to tell me that having the gateway on an uninterruptible power supply was the problem. Good heavens. I’ve been unable to get the issue escalated to knowledgeable staff.

We have Starlink as a alternate ISP and while there are occasional dropouts of 1 min or less presumably due to gaps in the satellite system or occasional reboots around 3-4am, it’s latency is much more stable than TMHI, as shown on this graph:

This seems to rule out our network from the router downstream as the source of the issue.

So the questions are: is this typical of the service everywhere?

It almost seems like a loose electrical connection in the tower equipment. My gateway is connected to the router via a new ethernet cable, which made no difference.

I only have one gateway so I’m unable to swap it out to completely rule it out. Rebooting it does not help. I’ve tried it many times.

Thanks for your time in helping with this.


6 replies

Userlevel 4
Badge +4

Maybe you could check to see if it’s correlated to a specific 5G band (N41, N71) or cell tower when you get the slowdown. I haven’t done a formal test, but I think my latency statistics are slower on N71 vs N41.  Still, latencies over 100 ms seem excessive, so maybe there is a congestion issue with a tower you are using.

Badge

Thanks; that’s an interesting idea. I’ll record current signal information and compare when the upset happens again. I’m not sure I know how to tell what tower the gateway is connecting with. If this turns out to be a factor, that would be great, except I still have little idea how to reach someone competent at T-Mobile with the information.

The service’s latency is outstanding when it’s not having an episode and it’s better than CenturyLink DSL, but that’s a really, really low bar to clear.

Userlevel 4
Badge +4

If you use the TM Internet App...under “Advanced Cellular Metrics”, you can see the tower ID (CGI) of the tower you are using….and also the 5G and 4GLTE bands you are connected to.

Badge

Thanks; found both on the app. Next will be to figure out how to map CGI to actual towers and confirm I’m connected to what I think I’m connected to.

Badge

A new episode began this morning. Latency was still reasonably low but the standard deviation of latency increased significantly.

I tested whether the Arcadyan gateway had changed towers since the baseline low-latency measurement on Jan 8, and I found it had.

I tested whether a re-boot would have it seek a different tower (gateway had been up over 90 days) and it did, and latency significantly improved immediately following the reboot.

Below is a comparison, l-r, of the baseline, the high-latency-stdev, and post-reboot. Post reboot, the gateway sought a third tower/antenna different from the baseline low latency and different from the increased latency tower/antenna, and on the third tower, latency returned to a low level and low stdev:

Latency graph showing effect of reboot that switched towers 1-hour scale:

Latency continues now to remain very low:

So apparently I can rectify the issue when it occurs with a gateway reboot, but this is far from optimal.

Question: why does the gateway switch towers or antennas to one with poorer performance, sometimes for days, and how do I prevent it?

Userlevel 4
Badge +4

Great detective work!   I notice when you switch towers, your bands may change also.  Sadly, the tower and band selections are outside of user control.  I’m sure there is a lot that goes into the algorithms that determine when to make those changes, but in may never make sense to us end-users :(

About the only thing I could suggest is to try relocating your router in your house to another location away from the tower that seems to have bad latency (if that is only one of them).  Also, I think TM has control over what towers you connect to...if you’re sure there is just one problematic tower, you may be able to get them to remove that ID from your access list.

Reply