Kitz Forum
Chat => Tech Chat => Topic started by: Weaver on May 03, 2018, 02:11:32 AM
-
A new record level of exaggeration. One speed tester, SpeedSmart, now claims that I have 1.86 Mbps of upstream capacity. I make this 43% greater than the possible true maximum. Other speed testers come up with 1.26 - 1.33 Mbps upstream. It's interesting that they are only so wild in the upstream figures though.
Anyone any idea what is going on?
As for downstream, the "speedof.me" website is consistently very flattering to downstream.
-
Hmm, not sure.
I tried it on my connection.
Speedtest.net
PING 6 ms
DOWNLOAD 114.69 Mbps
UPLOAD 23.07 Mbps
SpeedSmart
Ping 9 ms
Download 97.90 Mbps
Upload 22.82 Mbps
Speedof.me
Ping 8 ms
Download 123.48 Mbps
Upload 25.14 Mbps
DSLReports (single threaded)
Ping 8 ms
Downstream 116.3 Mbps
Upstream 24.89 Mbps
My upstream is a slightly sluggish this morning, typically was a few megabits higher yet there's minimal network activity here at the moment so god knows why that is lol.
-
So wildly different speed scale, ok parity between testers. Weird that my results are so very consistent (for me).
-
I have speedtest-cli (https://github.com/sivel/speedtest-cli) installed and this is what it shows for my circuit --
[Duo2 ~]$ speedtest-cli
Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from TalkTalk (WW.XX.YY.ZZ)...
Selecting best server based on latency...
Hosted by Ai Networks Ltd (Welwyn Garden City) [70.78 km]: 52.65 ms
Testing download speed........................................
Download: 4.86 Mbit/s
Testing upload speed..................................................
Upload: 0.83 Mbit/s
[Duo2 ~]$
-
Speedsmart does have an unusual (and bad) way of calculating the speed.
It shows a DS of 44 where my real max is 42. The other results are server related (I'm at least 3000km from those servers).
Speedtest.net used a server 100-200km away, so that result is the most accurate.
Speedsmart's France server, used for this test, was as fast as an overseas server could get, though.
Speedtest.net
Ping 10ms
DS 42.12
US 3.23
Speedsmart
Ping 81ms
DS 44.04
US 3.25
Speedof.me
Ping 69ms
DS 14.99
US 3.19
DSLReports (single threaded)
Ping 78ms
DS 27.5
US 2.56
-
@spring
When you say a server 100-200km away is that from where you live or from where your ISP connects to the internet? (A server I could walk to in 5 minutes could be 200 miles away by the route wires go from here to the server).
-
Upload speeds:
Speedsmart: 19.39 Mbps
Speedtest.net: 18.65 Mbps
Thinkbroadband: 18.8 Mbps
Speedof.me: 19.55 Mbps
DSLReports: 18.41 Mbps
BT Wholesale: 16.45 Mbps
-
At 1841 hours, today --
[Duo2 ~]$ speedtest-cli --simple
Ping: 57.011 ms
Download: 4.20 Mbit/s
Upload: 0.88 Mbit/s
[Duo2 ~]$
-
@spring
When you say a server 100-200km away is that from where you live or from where your ISP connects to the internet? (A server I could walk to in 5 minutes could be 200 miles away by the route wires go from here to the server).
ISP. How do I check my vdsl line ping?
Edit: I assume the best way is with the router, and using what I think is the only appropriate feature in its GUI:
traceroute to bezeqint.net (212.179.240.156), 30 hops max, 38 byte packets
1 bzq-179-37-1.cust.bezeqint.net (212.179.37.1) 10.544 ms 10.242 ms 9.899 ms
2 10.250.3.70 (10.250.3.70) 10.500 ms 9.868 ms 10.912 ms
3 bzq-25-77-22.cust.bezeqint.net (212.25.77.22) 29.941 ms 13.921 ms 10.613 ms
4 bzq-179-124-86.cust.bezeqint.net (212.179.124.86) 10.937 ms 11.532 ms 10.912 ms
5 bzq-25-88-241.cust.bezeqint.net (212.25.88.241) 172.486 ms 10.670 ms 10.583 ms
6 * * *
7 * * *
8 * * *
9 * * *
10 * * *
----Traceroute end----
Pinging default gateway:
PING 212.179.37.1 (212.179.37.1): 56 data bytes
64 bytes from 212.179.37.1: seq=0 ttl=255 time=10.292 ms
64 bytes from 212.179.37.1: seq=1 ttl=255 time=10.261 ms
64 bytes from 212.179.37.1: seq=2 ttl=255 time=10.396 ms
64 bytes from 212.179.37.1: seq=3 ttl=255 time=10.310 ms
--- 212.179.37.1 ping statistics ---
4 packets transmitted, 4 packets received, 0% packet loss
round-trip min/avg/max = 10.261/10.314/10.396 ms
Well, this isn't vdsl line ping because it's an IP but that gives you an idea how close the speedtest.net server was.
-
I decided that I ought to just get a big file from somewhere and download it and time it.
What's the easiest way to do it on an iPad with a straight TCP-based download? Something in a Web browser presumably.
I used a 50 'MB' (whatever that means) download test file (https://www.thinkbroadband.com/download) on the thinkbroadband website and downloaded that on my iPad in Safari and that took ~57s (with an error bound of 2 secs diameter) so I make that 7.02 Mbps downstream if we assume that they mean ‘MB’ = 106 = 1000000 bytes = 1 true MB in the correct notation as they didn't say MiB for 220.
I did a number of downloads, with 50MB and 100MB sizes and took the fastest result, as any slow-downs have to be deficiencies in the test set up, congestion, delays in the server, other traffic and so-on. I tried to get an estimate of the error due to TCP start time but failed due to variation in the results. I removed one-off delay due to DNS lookup by repeating tests so ensuring that DNS queries are cached.
This is to be compared with a downstream sum-of-IP-TX-rates of 7552566 bps reported by AA the ISP summed over the three lines and the far less meaningful 8591kbps is the sum-of-downstream sync rates. So that means TCP payload rate is 93% of the AA TX rate and since IPv6+TCP headers cost 4% then that isn't bad, TCP being better than 97% efficient compared to 100%=the maximum payload rate TCP could possibly achieve, calculated by knocking out TCP+IPv6 headers. It could have been IPv4, I didn't check, mea culpa. The truth will be a bit better than 93% because slow start was not removed from this.
It also means that (triple) multi line IP-bonding works fairly well, costing rather less than 4%, but otherwise quite unknown from these numbers.
-
I should have linked to it : https://www.thinkbroadband.com/download
Would anyone else care to try this and see if they get the same kind of efficiency ratios as I did? ie vs sync rate, or, better, vs sync rate less eg PPPoA or PPPoEoA overheads or PTM+VDSL2 overhead for those on FTTC, which we can help with (kitz quotes I think 88.2% for ADSL and I forget what two much higher numbers for VDSL2)
Also I would like to know how these straight-download numbers compare to the sometimes crazy stuff quoted by speed testers.
In all cases, we take the highest (best performance) numbers in a sequence of tests, although it is slightly interesting to see how much variation there is.
Ideally measurements should be obtained in the dead of night, but just watch out in case some kind of eg backup job might be running.
-
So the in-building phone wire that got cut isn't twisted anymore, it was gel crimped yesterday, which 99% of the benefit is so far preventing corrosion [the half-year corrosion made the tip fall off], and the phone wire is single strand from what appears to be cheap quality. Don't know if it matters and what single strand does for interference, for example from the other 20 phone wires curled into the same ball, but it's meaningless to me as I'm getting 150 FEC's per day and nothing more. And I'm still using 2012-ish generic adsl2 filters which I need to change xd.
PPPoE 1500 ECI
====================================================================================
VDSL Training Status: Showtime
Mode: VDSL2 Annex B
VDSL Profile: Profile 17a
G.Vector: Disable
Traffic Type: PTM Mode
Link Uptime: 1 day: 4 hours: 6 minutes
====================================================================================
VDSL Port Details Upstream Downstream
Line Rate: 3.658 Mbps 44.879 Mbps
Actual Net Data Rate: 3.455 Mbps 44.880 Mbps
Trellis Coding: ON ON
SNR Margin: 19.4 dB 25.3 dB
Actual Delay: 2 ms 0 ms
Transmit Power: -15.2 dBm -15.0 dBm
Receive Power: -17.5 dBm 6.7 dBm
Actual INP: 2.0 symbols 44.0 symbols
Total Attenuation: 0.0 dB 10.4 dB
Attainable Net Data Rate: 35.001 Mbps 110.044 Mbps
====================================================================================
VDSL Band Status U0 U1 U2 U3 U4 D1 D2 D3
Line Attenuation(dB): 0.1 10.0 15.0 N/A N/A 6.0 12.6 20.4
Signal Attenuation(dB): 0.1 9.9 14.9 N/A N/A 5.9 12.5 20.4
SNR Margin(dB): 24.4 20.1 18.9 N/A N/A 25.3 25.3 25.3
TX Power(dBm): -17.8 -19.3 -28.3 N/A N/A 12.3 8.1 6.7
====================================================================================(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww.speedtest.net%2Fresult%2F7329952919.png&hash=521530f0988d93e32a2fc93160052edeaf30f802) (http://www.speedtest.net/result/7329952919)
Edit: Bufferbloat without QoS was 300ms on dslreports, D score. After setting VMG1312-B10A QoS to Enabled with Upstream by Packet Length, without any rules, it's this (click image for details):
(https://forum.kitz.co.uk/proxy.php?request=http%3A%2F%2Fwww.dslreports.com%2Fspeedtest%2F33834061.png&hash=f2dc25fd719a9ed88c9bfd9c5417419a60685de1) (http://www.dslreports.com/speedtest/33834061)
Sites load noticeably faster too. Totally didn't expect that. QoS is godly :angel:
-
Speedsmart : upstream: 1.71 Mbps.
Ookla speed tester app: upstream: 1.10 Mbps
Which is correct? The sum of the sync rates is 1.44 Mbps and the real result can’t be that high unless Speedsmart is assuming things about overheads but it cannot know what type of DSL connection it is. Might not even be DSL at all in fact. So they can not be legitimately correcting for protocol overheads to quote a sync rate-like figure. The sum of the calculated IP throughout figures, derived by me from the sync rates, is 1.27 Mbps.
-
Speedsmart is simply incompetent.
-
I like this speedtester a lot (right now it's a bit buggy, but it's very nice): https://sourceforge.net/speedtest/
https://image.ibb.co/fAGepo/Untitled.png
-
Nice link, Spring. That speed tester reads impossibly high for my upstream too.