janvda / home

0 stars 0 forks source link

network problems 1st and 2nd floor #1

Open janvda opened 4 years ago

janvda commented 4 years ago

See also janvda/node-red-ping/issues :

janvda commented 4 years ago

Some statistics produced by flow janvda/node-red-network-test-client: from node pi3two which is connected via wifi to powerline adaptor tp_link1 (upstairs):

image

Observations

  1. A bit more than 1% of the time the network is not available
  2. About 2.5% of the time the network is not very responsive (average ping time = 300 ms).
  3. average ping time (when no problems) to nuc-jan is 8.9 ms

image

Observations

  1. The network unavailability is mostly due to the wifi connection between pi3two and tp_link1 !
  2. The high response times are also mostly due to the wifi connection between pi3two and tp_link1 !
  3. The average ping (when no problems) to tp_link1 is 5.2 ms
  4. So based on observation 3 an 6. The average ping time (when no problems) between tp_link1 and nuc_jan is 3.7 ms (= 8.9 ms - 5.2ms)
janvda commented 4 years ago

Some statistics produced by flow janvda/node-red-network-test-client from node nuc_jan where node pi3two (here it is called pi3two_wan) is still connected via wifi to powerline adaptor tp_link1 (upstairs)

image

Observations

  1. A bit less than 1% of the time the node pi3two is not accessible.
  2. About 2.8% of the time the node pi3two is not very responsive (average ping time = 423 ms).
  3. average ping time to pi3two_wan is 10.4 ms (this is close to the 8.9 ms that is measured in the opposite direction - see observation 3 in previous comment)
  4. So above observations are very similar as the observations seen in previous comment were the communication from pi3two_wan to nuc_jan is tested. ... they are not the same (differ a bit) most likely because the interval is a bit different (these observations started 1 hour later).

image

Observations

  1. The node tp_link1 was always accessible (unavailability = 0%) - this confirms observation 4 in previous comment.
  2. Only 0.7% of the time high ping times are seen - this confirms observation 5 in previous comment.
  3. The average ping time (when no problems) to tp_link1 is 4.2 ms (this is close to the 3.7 ms that is derived in observation 7 of previous comment).
janvda commented 4 years ago

Based on the previous 2 comments, we can conclude that the main network problem is the wifi connection between tp_link1 and pi3one.

janvda commented 4 years ago

The below chart reports the ping times from nuc_jan to pi3two_wlan and tp_link1. Note that a ping time = 2K means a time-out !

image

Observations

  1. The last 34 hours no ping time-outs are reported for tp_link1 - this is inline with above observations.
  2. The last 34 hours several ping time-outs and high response times are reported for tp_link1 - this is inline with above observations.
  3. before 7th of March there are periods with several ping timeouts to tp_link1. These things are not seen in above observations !.
janvda commented 4 years ago

I have moved the pi3two to 2nd floor (room garden side). Here below the statistics measured by both devices (pi3two and nuc_jan).

image

Observations for ping pi3two <=> nuc_jan (in both directions)

  1. About 0.85% of time the network was unavailable.
  2. About 3.5% of time the network was not very responsive (> 40ms)
  3. Average ping time when no problems is 10 ms.
  4. The main difference compared to pi3two on the 1st floor next to tp_link1 is that there are more periods that the network is not very responsive (3.5% compared to 2.5%).

Observations for pinging tp_link1 from nuc_jan

  1. An availability of 100% (this is inline with earlier observations as expected).
  2. Only 0.8% of the time big response times are seen (this is inline with earlier observations as expected).
  3. average ping time when no problems is 4.6 ms (this is inline with earlier observations as expected).

Observations for pinging tp_link1 from pi3two (FYI this is connected over wifi).

  1. About 0.92% not available - this is almost similar to the 0.87% that nuc_jan was not available (in fact I would not expect the unavailability % for tp_link1 higher than nuc_jan but that could be due to measurements errors).
  2. Average ping time when no errors is 6 ms.
  3. About 3% of the time big response times are seen.
janvda commented 4 years ago
  1. I have installed router tp-link archer C6 and connected it to tp_link1 (via ethernet). It is configured in bridge mode
  2. I have "Disabled" tp_link1 Wireless Radio so tp_link1 wifi cannot interfere with wifi of tp-link archer C6 router.
  3. Installed wifi-connect on pi3one so I can change wifi settings on this device.
janvda commented 4 years ago

setup:

Here below the ping stats coming from node pi3one based on tp-link archer C6 (see previous comment)

image

Here below the ping stats from node nuc-jan

image

janvda commented 4 years ago

just discovered that it might report a ping time = 5000. This doesn't seem a ping time but rather a time-out. So I have classified a ping time >= 5000 as unavailable.

janvda commented 4 years ago

I have moved tp_link1 and archer_c6 to the garage.

... _but at some point tp_link1 got disconnected from the powerline and to fix this I had to plug / unplug tplink1.

image

Observations

janvda commented 4 years ago

Here below the ping stats where

image

image

Observations

  1. 100% availability !!
  2. About 1.8% of the time high (> 40ms) response times were reported (on average 500ms).
  3. The high ping response times are due to the wifi connection between pi3two_wan and archer_c6.
  4. average ping time (when < 40ms) between nuc_jan and tp_link1 is 4.5 ms
  5. average derived ping time (when < 40 ms) between tp_link1 and archer_c6 is 0.07ms
  6. average derived ping time (when < 40 ms) between archer_c6 and pi3two_wan is about 5ms