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:

Pages: 1 ... 55 56 [57] 58 59 60

Author Topic: I have G.INP on ECI  (Read 231941 times)

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: I have G.INP on ECI
« Reply #840 on: July 15, 2016, 02:07:00 PM »

(Whoops --- I wrote this yesterday, and obviously forgot to post it)

I'll go check through the stats in a minute...

The portion of latency added by interleaving, when G.INP is activated, probably accounts for 0.15ms downstream plus 0.17ms upstream, making a total of 0.32ms.  Detailed calculations below.

That is approximately 3% of the extra latency of your mode 2 (interleaving when G.INP is disabled), when INP=2 and delay=8ms (down) and 3ms up.


Here are my calculations:

1. Downstream Mode 2
In mode 2, your modem would have to fill a block of data sized 837*128, total 107kbytes, before transmitting. The DSLAM would have to receive the whole block before de-interleaving. Processing that 107kbytes is what gives 8ms of latency.

2. Downstream Mode 3
In mode 3, the block is sized 8*254, total 2kbytes. If it takes 8ms to interleave/de-interleave 107k, then it takes 1/53th of the time to interleave/de-interleave just 2k ... 1/53 of 8ms is 0.15ms

3. Upstream Mode 2
In mode 2, the DSLAM has to fill a block of data sized 285*44, total 12kbytes, before transmitting. Your modem has to receive the whole block before de-interleaving. Processing that 12kbytes is what gives 3ms of latency.

4. Upstream Mode 3
In mode 3, the block is sized 8*82, total 0.7kbytes. If it takes 3ms to interleave/de-interleave 12k, then it takes 1/17th of the time to interleave/de-interleave just 0.7k ... 1/17 of 3ms is 0.17ms

NB: The calculation will be slightly wrong on the downstream side, because you likely synchronised at a higher speed in mode 3 (I estimate 101Mbps instead of 92Mbps), because the G.INP configuration also uses less overhead for FEC.

With the higher sync speeds, bits are travelling about 10% faster, so the block travels 10% quicker, and the latency is likely to be reduced about 10% - down to about 0.135ms
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: I have G.INP on ECI
« Reply #841 on: July 15, 2016, 03:54:42 PM »

it's not the ping it's interleaving the problem.

Interleaving only has two effects:
a) To alter the bit error distribution, to allow FEC to correct more errors
b) To alter the ping time, aka latency.

If it isn't the ping time, the latency, then the only other part that interleaving can affect is the error rate. But your line doesn't look to suffer many errors.

BUT...

G.INP has an extra impact on latency, that isn't caused by interleacing. If the retransmission mechanism is used after some packets are corrupted (and cannot be fixed by FEC), then there will be some extra latency for those packets only. That appears as jitter, where the latency varies a lot.

Again, your line doesn't seem to suffer from errors (FECs aren't important here), so I doubt this is the problem.

If you want me to confirm this, you'll need to provide two sets of statistics like last time, but separated over 24 hours.

things get a bit better when i change these things:

What I really meant was: How do you know that your problem is with G.INP if the only change you can make to get rid of G.INP makes things worse? How can you tell that fastpath is certain to be an improvement?

if performs worse than adsl in terms of latency and makes online gameplay like adsl with 7 mb down and 0.256 up 2006 times

When you moved to a "fiber" service, your backhaul probably changed too ... your gaming problems might be being caused there (for example, by sharing a congested link), and the latency of your G.INP access connection plays only a small part.

Did you move ISP when changing to fibre? Do you have any measured "ping" times from the command-line, rather than from within games?

For example, my round-trip ping time can be seen in "ping" commands to the BBC:
Quote
C:\>ping www.bbc.co.uk

Pinging www.bbc.net.uk [212.58.246.95] with 32 bytes of data:
Reply from 212.58.246.95: bytes=32 time=12ms TTL=55
Reply from 212.58.246.95: bytes=32 time=13ms TTL=55
Reply from 212.58.246.95: bytes=32 time=13ms TTL=55
Reply from 212.58.246.95: bytes=32 time=13ms TTL=55

Ping statistics for 212.58.246.95:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 12ms, Maximum = 13ms, Average = 12ms

When my line was fastpath, I'd get times of around 12-13ms. When my line was faulty, and I was given interleaving settings (INP=3, delay=8ms on downstream only), those 13ms times increased to 21ms. Now I'm on G.INP, and my ping times stayed the same: 12-13ms.

In fact, I have some good images to show those last numbers. Let me go and get them off a different computer...
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: I have G.INP on ECI
« Reply #842 on: July 15, 2016, 04:04:14 PM »

Finally, I have a couple of graphs to show you that give you an idea of what happens to latency:

a) When the line changes from "fastpath" (with a lot of errors) to "FEC + Interleaving" with INP=3, delay=8ms (with few errors)
This was my very first FTTC line in 2011, and the graph shows the ping round-trip latency.

I think you can readily see the impact of the 8ms extra latency. It is *very* obvious.

b) When the line changes from "fastpath" (with very few errors) to "G.INP" (with no errors).
I am part of the UK national "Ofcom" broadband monitoring test group, and my line has a test box permanently connected, running tests.

I have attached a graph that covers results for "UDP latency" (ie equivalent to ping round-trip time, but using UDP) from February 2015 through to July 2016.
At the start (February 2015), my line was "fastpath", and encountered very few errors.
On 27th March 2015, BT changed my line to G.INP for both upstream and downstream, as part of their rollout of G.INP nationwide.
On 26th June 2015, BT changed the line to put upstream back to fastpath, again as part of a nationwide rollout (too many modems failed with G.INP upstream).

The worst thing you can say is that the the line went slightly more unstable in the first month of deployment (I didn't notice any difference though).
But the latency times became indistinguishable from fastpath, and stayed that way.
Logged

willc

  • Member
  • **
  • Posts: 19
Re: I have G.INP on ECI
« Reply #843 on: July 15, 2016, 04:49:57 PM »

thanks for the answers.
so i'm waiting for the package xxxx,i'll do some tests and report (if everything goes by the plan)
today played on a german server with 24 of ping, and me butt got hurt,simply my bullets don't work,specially when somebody shoots at me.
by the way this is mine,will use for comparison:
« Last Edit: July 15, 2016, 05:06:30 PM by willc »
Logged

d2d4j

  • Kitizen
  • ****
  • Posts: 1103
Re: I have G.INP on ECI
« Reply #844 on: July 15, 2016, 05:25:44 PM »

Hi

Wwwombat, do you mean your a Sam knows tester or if not, I would be interested to become ofcom tester as well

Willc, your pings are high to bbc, what does pathping result show - open command box, pathping bbc.co.uk

Just thinking this may give a better clue and who is your ISP

Many thanks

John
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: I have G.INP on ECI
« Reply #845 on: July 15, 2016, 06:03:39 PM »

@d2d4j

Note that @willc is in Italy, so is likely to have long pings to the BBC servers in the UK. 40ms isn't that bad...

If I do the opposite, and try to ping "www.rai.it", I unfortunately reach some Akamai server located in the UK, so get a bogus reading.

If I try to "traceroute" to (for example), the Telecom Italia website, I find it takes me 12ms to get to London, 38ms to get to the Milan gateway on Seabone, and 47-50ms to get through to TIM's own network. Judging from the Seabone map, that extra 12ms is consistent with a hop from Milan to Rome.
Quote
Tracing route to tim.it [156.54.69.9]
over a maximum of 30 hops:

  1    <1 ms    <1 ms     1 ms  192.168.0.1
  2    11 ms    11 ms    11 ms  lo0.16.Central16.pcl-bng01.plus.net [195.166.130.210]
  3    17 ms    11 ms    11 ms  irb.15.pcl-cr01.plus.net [84.93.249.161]
  4    12 ms    12 ms    12 ms  195.99.126.96
  5    13 ms    12 ms    12 ms  core3-te0-10-0-19.faraday.ukcore.bt.net [109.159.249.23]
  6    13 ms    13 ms    13 ms  peer5-BE5.telehouse.ukcore.bt.net [213.121.193.171]
  7    14 ms    12 ms    13 ms  t2c4-et-8-1-0-0.uk-lon1.eu.bt.net [166.49.211.248]
  8    13 ms    12 ms    13 ms  ae14.londra32.lon.seabone.net [89.221.43.172]
  9    37 ms    37 ms    37 ms  ge6-1-8.milano1.mil.seabone.net [195.22.208.36]
 10    38 ms    38 ms    38 ms  ibs-resid.milano1.mil.seabone.net [93.186.128.210]
 11     *        *        *     Request timed out.
 12     *        *        *     Request timed out.
 13    50 ms    50 ms    50 ms  crs-mi003-vl4.opb.interbusiness.it [151.99.75.244]
 14     *        *        *     Request timed out.
 15     *        *        *     Request timed out.
 16    47 ms    47 ms    47 ms  80.21.4.83
 17    47 ms    47 ms    47 ms  host46-83-static.86-94-b.business.telecomitalia.it [94.86.83.46]
 18     *        *        *     Request timed out.
 19    49 ms    48 ms    49 ms  host117-38-static.77-62-b.business.telecomitalia.it [62.77.38.117]
 20    50 ms    48 ms    48 ms  host118-38-static.77-62-b.business.telecomitalia.it [62.77.38.118]
 21     *        *        *     Request timed out.
 22     *        *        *     Request timed out.
 23    49 ms    49 ms    50 ms  156.54.69.9

and, just to show that "ping" is consistent with "traceroute":
Quote
>ping www.tim.it

Pinging tim.it [156.54.69.9] with 32 bytes of data:
Reply from 156.54.69.9: bytes=32 time=49ms TTL=236
Reply from 156.54.69.9: bytes=32 time=49ms TTL=236
Reply from 156.54.69.9: bytes=32 time=49ms TTL=236
Reply from 156.54.69.9: bytes=32 time=49ms TTL=236

Ping statistics for 156.54.69.9:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 49ms, Maximum = 49ms, Average = 49ms
Logged

WWWombat

  • Kitizen
  • ****
  • Posts: 1674
Re: I have G.INP on ECI
« Reply #846 on: July 15, 2016, 06:09:39 PM »

Wwwombat, do you mean your a Sam knows tester or if not, I would be interested to become ofcom tester as well

Yes - that graph is the output from the tests run by my SamKnows whitebox for the Ofcom testing.

I've been running those since 2011, but they sent me a new box in February last year and I lost all my previous results.
Logged

d2d4j

  • Kitizen
  • ****
  • Posts: 1103
Re: I have G.INP on ECI
« Reply #847 on: July 15, 2016, 06:10:12 PM »

Hi wwwombat

Ah sorry, I did not realise or missed willc in Italy sorry.

Your correct that would explain the higher pings to bbc

Please ignore pathping as it would not help sorry

Many thanks

John
Logged

d2d4j

  • Kitizen
  • ****
  • Posts: 1103
Re: I have G.INP on ECI
« Reply #848 on: July 15, 2016, 06:11:31 PM »

Hi wwwombat

Many thanks, good to meet another Sam knows tester and mine has not been replaced since 2011 I think

Many thanks

John
Logged

willc

  • Member
  • **
  • Posts: 19
Re: I have G.INP on ECI
« Reply #849 on: July 18, 2016, 12:57:05 PM »

looks like a nice piece of kit
http://www.selta.com/ProductDetail?id=01t20000004m10XAAQ
look into the Asus DSL-AC68U, absolutely useless in the UK. might be better on an unlocked dslam.
i sent u some pm
Logged

j0hn

  • Kitizen
  • ****
  • Posts: 4104
Re: I have G.INP on ECI
« Reply #850 on: July 18, 2016, 04:56:47 PM »

nothing unexpected there. I don't think the "g.inp profile" was ever able to be selected by an ISP, it's all handled automatically by dlm.

sounds from that though that OpenReach are no longer "individually re-profiling" lines, at least fir sky, as some ISP's allow you to switch from standard to speed profile.
Logged
Talktalk FTTP 550/75 - Speedtest - BQM

deron

  • Member
  • **
  • Posts: 11
Re: I have G.INP on ECI
« Reply #851 on: August 04, 2016, 11:50:08 AM »

Is G.inp dead then? Or still planned...?
Logged

Dray

  • Kitizen
  • ****
  • Posts: 2361
Re: I have G.INP on ECI
« Reply #852 on: August 04, 2016, 11:53:03 AM »

On MyDslWebStats there is a user still showing G.INP on ECI
Logged

skyeci

  • Kitizen
  • ****
  • Posts: 1385
Re: I have G.INP on ECI
« Reply #853 on: August 04, 2016, 11:55:47 AM »

Is G.inp dead then? Or still planned...?

OR indicated a fix is being worked on but IF it returns,it  wont be till next year..
Logged

S.Stephenson

  • Reg Member
  • ***
  • Posts: 575
Re: I have G.INP on ECI
« Reply #854 on: August 04, 2016, 03:51:12 PM »

On MyDslWebStats there is a user still showing G.INP on ECI

I am one of the chosen few  :D

I have a feeling it's due to UHD BT TV as I took a TV package on my plusnet  line to see what would happen.
Logged
Pages: 1 ... 55 56 [57] 58 59 60
 

anything