Hi folks,
This is mostly to satisfy my curiosity, but if someone has technical insights I would be more than happy to hear. Background - yesterday my residential Sunrise fiber has changed the OTO plug from #1 to #2 (in Zurich). It's still XGS-PON, it's just that I'm using a different line now.
I have a setup that runs a continuous ICMP test towards various targets. I have observed that after changing the fiber plug, all the RTTs suddenly dropped. My guess (may be wrong though) is that with OTO#1 I was connected to a more complex PoP setup than with OTO#2 and this is what reduced the RTT.
For interested, here are 3 graphs presenting the drop. For every graph the left side is OTO#1 and the right is OTO#2. Monday morning was the time of changing the line:
1) https://i.ibb.co/qRL1Hyb/1.png 2) https://i.ibb.co/9ZL60KB/2.png 3) https://i.ibb.co/k2YJzJL/3.png
Cheers, Mat
Hello Mat
On 29.11.2022 10:41, Mat Kowalski via swinog wrote:
For interested, here are 3 graphs presenting the drop. For every graph the left side is OTO#1 and the right is OTO#2. Monday morning was the time of changing the line:
I do have a connection on OTO Plug 1 in Zürich, even with the EWZ provided Fiber/Ethernet bridge (4 Ethernet-Ports), with one port used for a line to Cyberlink (1 Gbit/s). External physical systems within Switzerland are mostly 2 - 2.5 ms with IPv4, IPv6 sometimes slightly more. The DNS at 8.8.8.8 / 2001:4860:4860::8888 has 1.8 ms. And the VM I have at ungleich.ch, has 4.7 ms (IPv4), and 5.2 ms (IPv6).
Almost a decade ago I had a MC2-based (copper) line with 10 Mbit/s from Cyberlink and I do remember that for systems within Switzerland it was also at ~3 ms. I also have a Sunrise/UPC Cable connection with IPv4 only, there the latency is ~21.5 ms and 25.4 ms (for the VM). But during the night (since around end of August) it is significantly higher, sometimes even above 50 ms. This is not the first time and it usually does improve again after a few month, e.g. when they split the cable segment into smaller areas.
So for me it seems that your ISP probably has the most influence for the latency of your line with their infrastructure and transit behind.
Best regards, Fabian
You sure that the change didn‘t happen from plug 2 to 1?!
Plug 2 is (mostly) using Swisscom‘s BBCS infrastructure, while plug 1 belongs to ewz and is used by other service providers (Disclaimer #1: this applies only to Zurich, other locations may differ; Disclaimer #2: Init7 is using both plug 1 and 2 in Zurich for Fiber7, but not BBCS).
The latency drop could explain if your connection is no longer on BBCS, as it usually adds about 5 to 10ms.
-- Fredy Künzler
Init7 (Switzerland) Ltd. Technoparkstrasse 5 CH-8406 Winterthur https://www.init7.net/
Am 29.11.2022 um 12:49 schrieb Fabian Wenk via swinog swinog@lists.swinog.ch:
Hello Mat
On 29.11.2022 10:41, Mat Kowalski via swinog wrote: For interested, here are 3 graphs presenting the drop. For every graph the left side is OTO#1 and the right is OTO#2. Monday morning was the time of changing the line:
I do have a connection on OTO Plug 1 in Zürich, even with the EWZ provided Fiber/Ethernet bridge (4 Ethernet-Ports), with one port used for a line to Cyberlink (1 Gbit/s). External physical systems within Switzerland are mostly 2 - 2.5 ms with IPv4, IPv6 sometimes slightly more. The DNS at 8.8.8.8 / 2001:4860:4860::8888 has 1.8 ms. And the VM I have at ungleich.ch, has 4.7 ms (IPv4), and 5.2 ms (IPv6).
Almost a decade ago I had a MC2-based (copper) line with 10 Mbit/s from Cyberlink and I do remember that for systems within Switzerland it was also at ~3 ms. I also have a Sunrise/UPC Cable connection with IPv4 only, there the latency is ~21.5 ms and 25.4 ms (for the VM). But during the night (since around end of August) it is significantly higher, sometimes even above 50 ms. This is not the first time and it usually does improve again after a few month, e.g. when they split the cable segment into smaller areas.
So for me it seems that your ISP probably has the most influence for the latency of your line with their infrastructure and transit behind.
Best regards, Fabian _______________________________________________ swinog mailing list -- swinog@lists.swinog.ch To unsubscribe send an email to swinog-leave@lists.swinog.ch
100% certainty, the *lower* RTT happens on the plug #2, the *higher* happens on the plug #1. Revdns is not very helpful in identifying the stack here as for both I see a similar entry:
* xdsl-*.adslplus.ch for the old connection (slow) * adsl-*.adslplus.ch for the new connection (fast)
I'm very curious about the direction of change here, given what Fredy says, that in Zurich #2 is BBCS. It would be really really surprising to discover that Swisscom BBCS is faster than EWZ. Init7's check gives me the following info about the OTO
{ "pop": "790HOT", "fiber7": true, "fiber_max_speed": 25000000, "planned": { "fiber7": null, "xgspon": null, "fiber7x": null }, "veto": false, "topologies": { "eth": true, "xgspon": true, "dsl": false } }
but apart from the fact that I can use P2P and P2MP in my connection, it does not tell me more. Of course if someone has access to any check to confirm the backend for #1 and #2, I'm happy to provide the OTO ID.
Cheers!
On 29/11/2022 15:04, Fredy Kuenzler via swinog wrote:
You sure that the change didn‘t happen from plug 2 to 1?!
Plug 2 is (mostly) using Swisscom‘s BBCS infrastructure, while plug 1 belongs to ewz and is used by other service providers (Disclaimer #1: this applies only to Zurich, other locations may differ; Disclaimer #2: Init7 is using both plug 1 and 2 in Zurich for Fiber7, but not BBCS).
The latency drop could explain if your connection is no longer on BBCS, as it usually adds about 5 to 10ms.
-- Fredy Künzler
Init7 (Switzerland) Ltd. Technoparkstrasse 5 CH-8406 Winterthur https://www.init7.net/
Am 29.11.2022 um 12:49 schrieb Fabian Wenk via swinogswinog@lists.swinog.ch:
Hello Mat
On 29.11.2022 10:41, Mat Kowalski via swinog wrote: For interested, here are 3 graphs presenting the drop. For every graph the left side is OTO#1 and the right is OTO#2. Monday morning was the time of changing the line: 1)https://i.ibb.co/qRL1Hyb/1.png https://i.ibb.co/qRL1Hyb/1.png
I do have a connection on OTO Plug 1 in Zürich, even with the EWZ provided Fiber/Ethernet bridge (4 Ethernet-Ports), with one port used for a line to Cyberlink (1 Gbit/s). External physical systems within Switzerland are mostly 2 - 2.5 ms with IPv4, IPv6 sometimes slightly more. The DNS at 8.8.8.8 / 2001:4860:4860::8888 has 1.8 ms. And the VM I have at ungleich.ch, has 4.7 ms (IPv4), and 5.2 ms (IPv6).
Almost a decade ago I had a MC2-based (copper) line with 10 Mbit/s from Cyberlink and I do remember that for systems within Switzerland it was also at ~3 ms. I also have a Sunrise/UPC Cable connection with IPv4 only, there the latency is ~21.5 ms and 25.4 ms (for the VM). But during the night (since around end of August) it is significantly higher, sometimes even above 50 ms. This is not the first time and it usually does improve again after a few month, e.g. when they split the cable segment into smaller areas.
So for me it seems that your ISP probably has the most influence for the latency of your line with their infrastructure and transit behind.
Best regards, Fabian _______________________________________________ swinog mailing list --swinog@lists.swinog.ch To unsubscribe send an email toswinog-leave@lists.swinog.ch
swinog mailing list --swinog@lists.swinog.ch To unsubscribe send an email toswinog-leave@lists.swinog.ch
-
Fabian Wenk -
Fredy Kuenzler -
Mat Kowalski