New Labs feature: Fusion Line Profile

Advanced feature discussion, beta programs and unsupported "Labs" features.
38 posts Page 3 of 4
by thulsa_doom » Wed Feb 05, 2014 10:37 am
dane wrote:
geogriffin wrote:
last night (i think) i got switched back from fastpath to interleaved without my intervention.. is it switching back to default in some cases like on resync?


Could be. We run a performance optimization (PO) process each night, and on circuits which are unstable the first step is to enable interleaving. So, it's possible that the expert system diagnosed your fastpath configuration as being unstable, and decided to begin the process of performance optimization.

This is exactly what happened. At 12:08 the system saw the circuit as unstable. At 1:00am it committed its first reprofile to get things back in line. We can manually disable profile optimization on this line, of course.
John Fitzgerald
Sonic Technical Support
by dane » Wed Feb 05, 2014 10:38 am
george.belle wrote:
On my bonded Fusion:
Annex A:
Line 1 is up and in sync at 1273kbps up and 22975kbps down.
Line 2 is up and in sync at 1273kbps up and 20477kbps down.

Annex M:
Line 1 is up and in sync at 2317kbps up and 19003kbps down.
Line 2 is up and in sync at 2320kbps up and 18163kbps down.


Thanks for sharing stats.

So you've got a choice between 43Mbps/2.5Mbps and 37Mbps/4.5Mbps for your circuit. I'd opt for the latter, in my opinion the downstream loss is worth the near doubling in upstream. 37Mbps will feel about the same as 43Mbps in day to day use, but the nearly double upstream will be really noticeable.

In regards to Fastpath versus Interleave, and the 10ms of latency versus 22ms you're observing - I'd stick with Interleave, it makes a big difference in circuit stability and re-transmits, which can effect throughput a lot. You might try some large downloads with Fastpath on & off and check how it compares, but I think the 12ms or so difference is likely to be a good trade-off for the stability and potential throughput gains that Interleave brings.

PS: Don't forget - we won't support circuits where customers have adjusted profile settings, so please do not contact support in regards to these settings or the behavior of your circuit when any adjustments from the default have been made.
Dane Jasper
Sonic
by dane » Wed Feb 05, 2014 10:39 am
thulsa_doom wrote:
This is exactly what happened. At 12:08 the system saw the circuit as unstable. At 1:00am it committed its first reprofile to get things back in line. We can manually disable profile optimization on this line, of course.


Yup, makes sense, thanks for confirming. I've asked systems if they can exempt a circuit from PO while customer-configured Fastpath is in effect.
Dane Jasper
Sonic
by geogriffin » Wed Feb 05, 2014 10:56 am
how can i tell if my line is unstable in fastpath? i didnt and dont currently see any errors under "ATM traffic" or "IP traffic"... or any "excessive impluse noise." though maybe errors could have happened last night before or while the line was being (stress) tested; and then when my line got automatically reprofiled, the stats went away?
by desertflyer » Mon Mar 03, 2014 12:25 pm
In the sync rate graphs, what do gaps in the line mean? I don't think that I lost sync.

Image
by kgc » Mon Mar 03, 2014 3:09 pm
Most likely just ghosts in the machine relating to stats collection on such a huge number of objects.
Kelsey Cummings
System Architect, Sonic.net, Inc.
by digitalbitstream » Tue Mar 04, 2014 3:38 pm
Do gaps in the graph indicate a real DSL level glitch? And will a tiny short gap always show visually?

597.png
597.png (17.83 KiB) Viewed 79777 times
by tovbinm » Mon Mar 17, 2014 8:25 pm
My tests:

DOWNLOAD PRIORITY:
Download Speed: 19791 kbps (2473.9 KB/sec transfer rate)
Upload Speed: 1112 kbps (139 KB/sec transfer rate)
Latency: 31 ms
Uncanceled Echo: -17.2 dB

UPLOAD PRIORITY:
Download Speed: 14601 kbps (1825.1 KB/sec transfer rate)
Upload Speed: 2186 kbps (273.3 KB/sec transfer rate)
Latency: 22 ms
Uncanceled Echo: 15.6 dB
by rcsheets » Fri Apr 18, 2014 2:35 am
Is it possible to get the data being used to generate the graphs? In particular I'd like to be able to look back further than 24 hours on the bandwidth utilization graph, but without Sonic having to store more historical data. Something like a URL to poll to get the available data would be a neat feature.
by kgc » Fri Apr 18, 2014 9:31 am
Sorry, it's not practical for us to export this information to users. That said, most CPE probably provides similar information via the GUI if not over SNMP as well and it should be relatively straight forward to integrate that into a standard stats package.
Kelsey Cummings
System Architect, Sonic.net, Inc.
38 posts Page 3 of 4

Who is online

In total there are 3 users online :: 0 registered, 0 hidden and 3 guests (based on users active over the past 5 minutes)
Most users ever online was 999 on Mon May 10, 2021 1:02 am

Users browsing this forum: No registered users and 3 guests