Availability report accuracy
Hi, mostly out of curiosity: what sort of accuracy are the numbers in the monthly availability report? For example: the report says that my probe was "connected" 99.95% of the time (down for 20 seconds) in December. What does that mean? That the SSH control connection was not up? That the probe didn't have link? That it couldn't reach any targets? Something else? Cheers, Lorenzo
Hi all, I want to jump into this topic too, but before a Happy New Year to everybody. Was wondering too, for December my probe got an total availability of only 98.17% - and which is in fact not true. I'm on a VDSL line 50 mbit from German Telekom, and every night my router has to re-establish the connection to get a new public IP. According to the log files from my router this whole process takes only a couple of seconds, but in the Monthly probe report from atlas the time between Disconnected and new Connected is around 20 to 30 minutes, see examples below: +---------------------+---------------------+------------+--------------+ | Connected (UTC) | Disconnected (UTC) | Connected | Disconnected | |---------------------+---------------------+------------+--------------+ | 2015-12-26 02:35:51 | 2015-12-27 02:06:06 | 0d 23:30 | 0d 00:27 | | 2015-12-27 02:31:20 | 2015-12-28 02:04:36 | 0d 23:33 | 0d 00:25 | | 2015-12-28 02:39:09 | 2015-12-29 02:04:29 | 0d 23:25 | 0d 00:34 | | 2015-12-29 02:31:40 | 2015-12-30 02:03:09 | 0d 23:31 | 0d 00:27 | And this reports my router (times in CET) 27.12.15 03:07:12 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 27.12.15 03:07:15 Internetverbindung wurde getrennt. 27.12.15 03:07:16 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.136.176, ... 28.12.15 03:06:02 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 28.12.15 03:06:06 Internetverbindung wurde getrennt. 28.12.15 03:06:08 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.141.126, ... 29.12.15 03:04:46 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 29.12.15 03:04:49 Internetverbindung wurde getrennt. 29.12.15 03:04:50 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.145.147, ... How comes this huge difference? Cheers, Gerhard
Am 02.01.2016 um 18:27 schrieb Lorenzo Colitti <lorenzo@google.com>:
Hi,
mostly out of curiosity: what sort of accuracy are the numbers in the monthly availability report? For example: the report says that my probe was "connected" 99.95% of the time (down for 20 seconds) in December.
What does that mean? That the SSH control connection was not up? That the probe didn't have link? That it couldn't reach any targets? Something else?
Cheers, Lorenzo
it would be great to have a virtual probe so i could give it two fttc ethernet connections to overcome ip dhcp forced renew colin Sent from my iPhone
On 2 Jan 2016, at 17:57, Gerhard Joch <elektronik@normalzeit.org> wrote:
Hi all,
I want to jump into this topic too, but before a Happy New Year to everybody.
Was wondering too, for December my probe got an total availability of only 98.17% - and which is in fact not true. I'm on a VDSL line 50 mbit from German Telekom, and every night my router has to re-establish the connection to get a new public IP. According to the log files from my router this whole process takes only a couple of seconds, but in the Monthly probe report from atlas the time between Disconnected and new Connected is around 20 to 30 minutes, see examples below:
+---------------------+---------------------+------------+--------------+ | Connected (UTC) | Disconnected (UTC) | Connected | Disconnected | |---------------------+---------------------+------------+--------------+ | 2015-12-26 02:35:51 | 2015-12-27 02:06:06 | 0d 23:30 | 0d 00:27 | | 2015-12-27 02:31:20 | 2015-12-28 02:04:36 | 0d 23:33 | 0d 00:25 | | 2015-12-28 02:39:09 | 2015-12-29 02:04:29 | 0d 23:25 | 0d 00:34 | | 2015-12-29 02:31:40 | 2015-12-30 02:03:09 | 0d 23:31 | 0d 00:27 |
And this reports my router (times in CET)
27.12.15 03:07:12 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 27.12.15 03:07:15 Internetverbindung wurde getrennt. 27.12.15 03:07:16 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.136.176, ... 28.12.15 03:06:02 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 28.12.15 03:06:06 Internetverbindung wurde getrennt. 28.12.15 03:06:08 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.141.126, ... 29.12.15 03:04:46 Die Internetverbindung wird kurz unterbrochen, um der Zwangstrennung durch den Anbieter zuvorzukommen. 29.12.15 03:04:49 Internetverbindung wurde getrennt. 29.12.15 03:04:50 Internetverbindung wurde erfolgreich hergestellt. IP-Adresse: 87.158.145.147, ...
How comes this huge difference?
Cheers,
Gerhard
Am 02.01.2016 um 18:27 schrieb Lorenzo Colitti <lorenzo@google.com>:
Hi,
mostly out of curiosity: what sort of accuracy are the numbers in the monthly availability report? For example: the report says that my probe was "connected" 99.95% of the time (down for 20 seconds) in December.
What does that mean? That the SSH control connection was not up? That the probe didn't have link? That it couldn't reach any targets? Something else?
Cheers, Lorenzo
On 2016/01/02 18:57 , Gerhard Joch wrote:
I want to jump into this topic too, but before a Happy New Year to everybody.
Was wondering too, for December my probe got an total availability of only 98.17% - and which is in fact not true. I'm on a VDSL line 50 mbit from German Telekom, and every night my router has to re-establish the connection to get a new public IP. According to the log files from my router this whole process takes only a couple of seconds, but in the Monthly probe report from atlas the time between Disconnected and new Connected is around 20 to 30 minutes, see examples below:
+---------------------+---------------------+------------+--------------+ | Connected (UTC) | Disconnected (UTC) | Connected | Disconnected | |---------------------+---------------------+------------+--------------+ | 2015-12-26 02:35:51 | 2015-12-27 02:06:06 | 0d 23:30 | 0d 00:27 | | 2015-12-27 02:31:20 | 2015-12-28 02:04:36 | 0d 23:33 | 0d 00:25 | | 2015-12-28 02:39:09 | 2015-12-29 02:04:29 | 0d 23:25 | 0d 00:34 | | 2015-12-29 02:31:40 | 2015-12-30 02:03:09 | 0d 23:31 | 0d 00:27 |
Hi, I'm not completely sure, but I think the problem is that the probe is waiting for a TCP connection timeout. I guess that for some reason the probe doesn't get a TCP RST. Might be a firewall problem somewhere. Philip
On 2016/01/02 18:27 , Lorenzo Colitti wrote:
mostly out of curiosity: what sort of accuracy are the numbers in the monthly availability report? For example: the report says that my probe was "connected" 99.95% of the time (down for 20 seconds) in December.
What does that mean? That the SSH control connection was not up? That the probe didn't have link? That it couldn't reach any targets? Something else?
The measurement is about the ssh connection.
participants (4)
-
Colin Johnston -
Gerhard Joch -
Lorenzo Colitti -
Philip Homburg