Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Author Topic: Strange ‘change’ events in 3G  (Read 613 times)

Weaver

  • Senior Kitizen
  • ******
  • Posts: 11459
  • Retd s/w dev; A&A; 4x7km ADSL2 lines; Firebrick
Strange ‘change’ events in 3G
« on: January 10, 2020, 08:30:23 AM »

I have a 3G USB ‘dongle’ NIC plugged into my Firebrick router. My ISP is Andrews and Arnold (AA). At 15:45 approx yesterday (Thu 2020-01-09T15:45) there was a minute or so where PPP ping-packet loss was evident on the AA constant quality monitoring (CQM) service channel that monitors this device with continual PPP LCP echo requests/responses (‘ppp pings’). Before this event AA CQM reported a constant latency / RTT of 25 ms minimum and after the event this figure dropped to a long term constant 16 ms.

Important to note that a similar event happened at 23:10 but without the change in latency; latency was still low, unchanged at 16ms still, constant throughout the rest of the day. Severe packet loss, the same if anything, and several minutes long, same duration. So an exactly similar event in two respects: duration and peak depth, but without the lasting effects.

I’m trying to think what the significance of this time of day is. I thought of comparing it with sunset.

Siri tells me sunset is at 16:04, but the Western mountains of the Cuilfhionn mountain range are in the way. Siri has given 16:04 as a time calculated for Broadford in Skye so it should have the right allowance for the western longitude and the northern latitude but the mountains Blà-Bheinn and Beinn na Cailliche will take - I don’t know - five minutes or so maybe more off the effective sunset time at distance that the base station’s location on the east side of the valley below me.

At any rate, the sun will be low. I don’t know how that would explain the possible correlation would explain the time. I don’t understand what might determine the start and end of the packet loss period, a minute or two in duration. It seems clear to me that it is a ‘change of state event’ - there is a ‘before’ state #1 and an ‘after’ state #2 and it seems that the link is very unhappy for a while in state #1 at that time and it transitions to state #2. I don’t know if the ‘unhappiness’ is a one-off thing which forced a transition and whose effects were resolved by the change in state or whether it was the end if a period of deteriorating conditions which forced a change - speculating again.

The 23:10 packet loss event clearly is nothing to do with sunset at this time of year and it had no lasting effects, so this poses difficulties for theorising. This event suggests something external that starts, happens for a few minutes and then ends and the observed packet loss is caused by this. It would be a sort-of one-off in the sense that it is not regular, repeating, or paired and may or may not be easily or visibly ‘predictable’ in terms of external driver-conditions.

What does it mean to change to have a much lower latency like this in 3G?

BTW The link/service is capable of 4G/3G but the rubbish hardware only speaks 3G for some incredible reason.

Is it possible that this USB NIC is really running in LTE and changing to 3G, dropping down as that helps it cope with poor conditions? I don’t know if the device could have been incorrect described to me, underselling it a lot if so! I think my iPad changes ‘down’ to 3G when it is struggling in poor signal conditions. I don’t know if, or if so, why, 3G might possibly be more robust than LTE? That doesn’t help us understand the event at 23:11?

This thing happens several times in a week. On Wednesday at ~05:50 and again at ~07:25 there were two events of the second type - no latency change, but unlike the ~15:45 yesterday (Thu), there was no change of latency; nite that the latency was at the high #1 25ms constant level, constant pretty long term, several days.

Events over the last few days, since power-on:



Downwards-going red spikes are packet loss indications, height is the fraction of loss. The blue-yellow band at the bottom is latency, dark blue minimum latency, yellow maximum, bright blue is mean. See AA CQM graphs explained

The start of the graph at ~21:10 is when the kit was powered up, so the initial packet loss spike is presumably associated with that.
« Last Edit: January 10, 2020, 09:47:56 AM by Weaver »
Logged
 

anything