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: Upstream improvement?  (Read 396 times)

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 6376
  • Retd sw dev; A&A; 4 ◊ 7km ADSL2; IPv6; Firebrick
Upstream improvement?
« on: June 16, 2018, 02:10:28 AM »

Anything I can do to improve my upstream? Has been reduced by more than 100k (8%) since I switched modems to the ZyXEL VMG 1312-B10As from the DLink DSL-320B-Z1s becfore.
Logged

burakkucat

  • Global Moderator
  • Senior Kitizen
  • *
  • Posts: 24202
  • Over the Rainbow Bridge
    • The ELRepo Project
Re: Upstream improvement?
« Reply #1 on: June 16, 2018, 08:31:42 PM »

I am not aware of anything that could be tried. I presume you have configured QoS for the upstream link?
Logged
:cat:  100% Linux and, previously, Unix. Co-founder of the ELRepo Project.

Please consider making a donation to support the running of this site.

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 6376
  • Retd sw dev; A&A; 4 ◊ 7km ADSL2; IPv6; Firebrick
Re: Upstream improvement?
« Reply #2 on: June 16, 2018, 10:47:00 PM »

Wish I had such a thing. That is one glaring omission in the Firebrick, QoS.  I am told it is supposed to prioritise small packets but I am not completely convinced. Round trip times go up from ~40 ms to ~230 ms when a flat-out download is in progress.

But I was wondering if there is anything physical or otherwise that I could do to get more throughput in the upstream direction so that Facetime would work better.
Logged

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 6376
  • Retd sw dev; A&A; 4 ◊ 7km ADSL2; IPv6; Firebrick
Re: Upstream improvement?
« Reply #3 on: June 17, 2018, 12:17:10 AM »

I wonder if some sage could tweak the upstream SNRM target to bring it down below 6dB a bit ? Since I run ok at 3 dB downstream SNRM, trying a 3-4dB upstream target SNRM would be well worth a go.
Logged

burakkucat

  • Global Moderator
  • Senior Kitizen
  • *
  • Posts: 24202
  • Over the Rainbow Bridge
    • The ELRepo Project
Re: Upstream improvement?
« Reply #4 on: June 17, 2018, 01:00:03 AM »

I do not know of any method to manipulate the US SNRM target.
Logged
:cat:  100% Linux and, previously, Unix. Co-founder of the ELRepo Project.

Please consider making a donation to support the running of this site.

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 6376
  • Retd sw dev; A&A; 4 ◊ 7km ADSL2; IPv6; Firebrick
Re: Upstream improvement?
« Reply #5 on: June 17, 2018, 02:13:27 AM »

No, I was wondering about an actual bit of software surgery. And also if there might be anything physical that I can do to help very low frequencies.
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5441
Re: Upstream improvement?
« Reply #6 on: June 17, 2018, 04:31:43 AM »

Wish I had such a thing. That is one glaring omission in the Firebrick, QoS.  I am told it is supposed to prioritise small packets but I am not completely convinced. Round trip times go up from ~40 ms to ~230 ms when a flat-out download is in progress.

But I was wondering if there is anything physical or otherwise that I could do to get more throughput in the upstream direction so that Facetime would work better.
download is harder to shape and it gets harder the lower the burst speed, it works by manipulating the sending server to shrink its congestion window so it doesnt saturate downstream.
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 6376
  • Retd sw dev; A&A; 4 ◊ 7km ADSL2; IPv6; Firebrick
Re: Upstream improvement?
« Reply #7 on: June 17, 2018, 05:07:47 AM »

The thing is they tend to have big Fibrebricks 6000 series at the AA end so they could do QoS and shaping properly with those if someone out in the time to write the code. I know they have some Cisco kit nowadays too because they were struggling with it and cursing it at one point.

If Firebrick Ltd develops a 10Gbps or even 100Gbps switch or router or whatever then for all I know the Cisco stuff might get the boot so that AA eats their own dog food 100%.

I need to bribe our friend RevK somehow with something nice to get him to look at anti-bufferbloat active queue management (fq_codel or whatever itís called), then heuristic prioritisation and then even QoS.

How could I test small-packet heuristic prioritisation now ? I want to see if it is really working or not. I get the feeling that it isnít working but I donít know what I am talking about.
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5441
Re: Upstream improvement?
« Reply #8 on: June 18, 2018, 09:15:12 AM »

Well there is a second way, but its frowned up on, isps like plusnet when they used to heavily shape things like torrents, they actually went into each packet, and modified the TCP window size to something tiny to get sending speeds low, deep packet inspection form of traffic shaping, units like the firebrick, pfsense etc. dont use DPI.

The more polite way of doing it is simply moving the downstream bottleneck to the firewall device, this is achieved by making it smaller than the speed capability of the connection, then when throughput is higher than desired for queue management packets will get dropped, this then makes the TCP congestion window shrink in response which means the sender will send slower, its not aggressive so the reaction is delayed, and works less reliable the more download threads there is and the lower the burst speed is, its also more difficult on low RTT servers as they can ramp up speed so fast. So e.g. if you trying to tame steam, you want to choose a download server on the other side of the world, that alone will help noticeably.

Remember its the sender that sends the packets, so shaping is by far easier on outbound traffic than inbound, some people claim its impossible to shape inbound traffic, I disagree with that, but I acknowledge its harder to do it.
« Last Edit: June 18, 2018, 09:21:50 AM by Chrysalis »
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

spring

  • Reg Member
  • ***
  • Posts: 322
Re: Upstream improvement?
« Reply #9 on: June 18, 2018, 09:19:23 AM »

i wonder if they can run SQM at the ISP [download].
upload is always going to start at your LAN, then move to layer-3.

Edit:
Quote
Examples of deployed Smart Queue Management systems include CeroWrtís SQM implementation, OpenWrtís qos-scripts, IPFire, the Gargoyle router project, and Streamboost . WRED is deployed in many locations. France Telecom deploys SFQ. Free.fr has the first known large-scale fq_codel deployment, using three bands of fq_codel for tiers of QoS. (it is also the largest ECN enabled end-user deployment). The Streamboost product (now coming available in multiple 802.11ac routers) combines a bandwidth sensor, with a packet classification engine, with a multi-band fq_codel implementation.


about upstream target margin: https://forum.kitz.co.uk/index.php?topic=18254.0


also, this has been said about shielding:
Quote from: A different forum
Any shield (grounded or not) will add to the capacitive load on the line.
The problem with shielded cable it's not grounded the shielding in that cable acts like an Ariel picking up unwanted stuff from who knows where.
best is short twisted "unshielded" cables.

pure sync speed suggests untwisted:
I too found the Tandy cable synced slower than a bog standard ADSL cable. I have a hardwired VDSL extension through Cat5e and to that I have plugged in a Billion 8800NL cable and no issues.

http://forums.thinkbroadband.com/general/4562127-modem-cable-experiments-on-my-line.html#Post4562171

question is where to get the 7800/8800 or a different, similar cable with a length of 0.3-0.5m to try out. or twisted pair.
« Last Edit: June 18, 2018, 11:45:43 AM by spring »
Logged
No one knows what is the taste of the void.