Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[lalof72845]

Hello All,

I currently have VDSL2 (FTTN) and latency overhead for modulation was around 4ms on initial connect but after some time would jump to 8ms using DVA-2800. I have also tested BiPAC 8900X R3 (8-10ms) and DSL-AX82U (5-8ms). As the DSL-AX82U has a number of configration options, just wondering if you can suggest any settings that may improve the latency or keep it matching the initial connect latency? Happy to reduce bandwidth for better latency.

DSL-AX82U xdslctl command options on current firmware.

Code: [Select]

Usage: xdslctl start [--up] <configure command options>
       xdslctl stop
       xdslctl connection [--up] [--down] [--loopback] [--reverb]
           [--medley] [--noretrain] [--L3] [--diagmode] [--L0]
           [--tones <r1-r2,r3-r4,...>] [--normal] [--freezeReverb] [--freezeMedley]
       xdslctl configure/configure1 [--mod <a|d|l|t|2|p|e|m|M3|M5|v|r>] [--lpair <(i)nner|(o)uter>] [--phycfg1 param_name=0xmask/0xval]
           [--trellis <on|off>] [--snr <snrQ4>] [--bitswap <on|off>] [--sesdrop <on|off>]
           [--sra <on|off>] [--CoMinMgn <on|off>] [--minINP <sym>] [--maxDelay <ms>] [--i24k <on|off>] [--phyReXmt <0xBitMap-UsDs>]
           [--Ginp <0xBitMap-UsDs>] [--TpsTc <0xBitMap-AvPvAaPa>] [--monitorTone <on|off>]
           [--profile <0x00 - 0x1FF> | <"8a |8b |8c |8d |12a |12b |17a |30a |35b">] [--us0 <on|off>]
           [--dynamicD <on|off>] [--dynamicF <on|off>] [--SOS <on|off>] [--maxDataRate <maxDsDataRateKbps maxUsDataRateKbps maxAggrDataRateKbps>]
           [--forceJ43 <on|off>] [--toggleJ43B43 <on|off>]
       xdslctl bert [--start <#seconds>] [--stop] [--show]
       xdslctl afelb [--time <sec>] [--tones] [--signal <1/2/8>]
       xdslctl qlnmntr [--time <sec>] [--freq <msec>]
       xdslctl inm [--start <INMIATO> <INMIATS><INMCC><INM_INPEQ_MODE><INM_INPEQ_FORMAT>] [--show]
       xdslctl snrclamp [--shape <shapeId>] [--bpshape [bpIndex-bpLevel,]]
       xdslctl nlnm [--show ] [--setThld <Thld_Num_Tones>]
       xdslctl diag [--logstart <nBytes>] [--logpause] [--logstop] [--loguntilbufferfull <nBytes>] [--loguntilretrain <nBytes>] [--dumpBuf <sizeKb>]
       xdslctl ntr [--start [output freq(default is 8000)]] [--stop]
       xdslctl info [--state] [--show] [--stats] [--SNR] [--SNRM] [--QLN] [--Hlog] [--Hlin] [--HlinS] [--Bits]
           [--pbParams] [--linediag] [--linediag1] [--UER]
           [--TOD] [--reset] [--vendor] [--cfg] [--cfg-verbose] [--cfg-help]
       xdslctl profile [--show] [--save] [--restore]
       xdslctl --version
       xdslctl --help

[g3uiss]

I do think you mean FTTC. The quoted latency is very low! The normal expected latency with no interleave is 8ms.

[kitz]

If you discount any interleaving, then the first external hop will depend on your location.   Most ISP's interconnects are in one of the London co-los.  If you live in Manchester, then your first hop is going to be slightly higher than someone living in London.  If you live in Northern Scotland its going to be even higher.  There are a few exceptions such as if you are in Manchester and your ISP is Zen who have a gateway in Manchester....  but thats academic as you'll likely end up with your data having to go down to London most of the time anyhow.

As g3uiss said, 8ms is considered good for a first hop.

[Alex Atkin UK]

There are a few exceptions such as if you are in Manchester and your ISP is Zen who have a gateway in Manchester....  but thats academic as you'll likely end up with your data having to go down to London most of the time anyhow.

That's curious as routing to Manchester from Sheffield was always slower than using a London gateway, as it merely seemed to add another hop to land on the same gateway.  Of course its impossible to understand exactly what was going on given traceroute doesn't tell you how the backhaul routing is working.

[kitz]

In theory it should be faster, but we dont know the routing between Sheffield and Manchester.  It wont be straight as a bird flies.
 
Many years ago when we were on the old BT RAS and you could do a testlogin to the RAS... I picked up some horrendous frst hop latency that added about another 12ms.   It was almost like my routing went up to Edinburgh and then down to Manchester RAS.   It latest for about 3 weeks making first hop to London at about 28ms FAST mode.

Much of the first hop is taken actually getting on to the backhauls.  For example it could take 6ms to get to your exchange and then only 2ms to go hundreds of miles down to London.   Most of the routing is tunnelled so you cant actually see most of the hops between A&B.

[XGS_Is_On]

If you're already on fast path which you evidently are there's nothing you can configure.

[lalof72845]

I do think you mean FTTC. The quoted latency is very low! The normal expected latency with no interleave is 8ms.

Apologies, i am on FTTN but was on FTTP at my old house.

[lalof72845]

If you discount any interleaving, then the first external hop will depend on your location.   Most ISP's interconnects are in one of the London co-los.  If you live in Manchester, then your first hop is going to be slightly higher than someone living in London.  If you live in Northern Scotland its going to be even higher.  There are a few exceptions such as if you are in Manchester and your ISP is Zen who have a gateway in Manchester....  but thats academic as you'll likely end up with your data having to go down to London most of the time anyhow.

As g3uiss said, 8ms is considered good for a first hop.

The connection starts off around 4.x MS and then increases to 8ms. Then it should be possible to some how keep the lower latency, happy to give up the extra bandwidth. It might be related to interleaving or another related feature.

Below is a comparison of the connection state at 4ms and 8ms.
https://www.textcompare.org/?id=64d48a01ba114ff65286c9bf

I did find "0x00004000 CfgFlagsDisableVectoring" which if i am not mistaken is related to fast path instead of interleaving? I suspect that this setting can be set using "xdslctl start [--up] <configure command options>" but not sure the best way to get the current options used and what offset would be required.

[lalof72845]

If you're already on fast path which you evidently are there's nothing you can configure.

I don’t think I am on VDSL2 fast path right now but interleaving via G.998.4.

[Alex Atkin UK]

Which is g.INP right?  Which as I understand it does not incur a latency penalty as it merely stores a buffer on the DSLAM so it can retransmit corrupt data, rather than actually add overhead to correct it at the end-user end.

When you mentioned the latency in the first post, are you referring to additional latency or your actual first-hop ping time?

[lalof72845]

Which is g.INP right?  Which as I understand it does not incur a latency penalty as it merely stores a buffer on the DSLAM so it can retransmit corrupt data, rather than actually add overhead to correct it at the end-user end.

When you mentioned the latency in the first post, are you referring to additional latency or your actual first-hop ping time?

Correct G.INP that should not add latency in most cases but additional latency is occurring for some reason.

The latency between the modem and the ISP next hop route (E.g. Latency for modulation only not routing or extra hops). On a fresh L2 connection the connection is 4ms but increases to 8ms over time. I suspect it something similar to interleaving as i notice the sync speed also increases by a small amount (example provided in the compare link above)

[g3uiss]

Adding additional to what ? Your figures are really good. It’s been said 8ms is what you should expect. I guess you keep swapping devices, you get a different route each time. Nothing to do with the devices

[lalof72845]

Adding additional to what ? Your figures are really good. It’s been said 8ms is what you should expect. I guess you keep swapping devices, you get a different route each time. Nothing to do with the devices

I can confirm the latency is related to the modulation and not routing. This is first hop latency from the modem and not the total latency on a path. From what i have read up it is related to interleaving and G.INP.

Looking at other people that dont have this issue, they are using fast mode.

Path Mode Fast Fast
Interleave Depth 1 1

using "xdslctl configure --maxDataRate" i am getting close to fast path mode for some number of hours but would like to force keep it in this mode. I believe this will be related to either "0x00004000 CfgFlagsDisableVectoring" or forcing max depth to 1 just not sure on the possible commands.

[tubaman]

First hop latency of under 8ms is certainly possible as per tracert below from my connection (VDSL 42/7 G.INP Retx High), not that I know why such a small difference would be of any concern. 

Code: [Select]

tracert 8.8.8.8

Tracing route to dns.google [8.8.8.8]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     5 ms     5 ms     5 ms  172.16.14.60
  3     *        *        *     Request timed out.
  4     9 ms     8 ms     8 ms  62.172.102.70
  5    10 ms    10 ms    10 ms  peer3-et3-0-6.redbus.ukcore.bt.net [62.172.103.155]
  6    11 ms    11 ms    10 ms  109.159.253.3
  7    11 ms    10 ms    10 ms  216.239.54.127
  8     9 ms     9 ms     9 ms  108.170.234.231
  9    10 ms     9 ms     9 ms  dns.google [8.8.8.8]

Trace complete.

[Alex Atkin UK]

Lower is better but especially with FTTC its kinda nuts to be chasing such low numbers that frankly I've never heard of anyone achieving.

I can confirm the latency is related to the modulation and not routing. This is first hop latency from the modem and not the total latency on a path.

First hop latency doesn't prove its not related to routing, as how the actual connection is being tunneled across the network to get to your ISP can vary.

Even on fibre my connection can vary from 6-8ms and 4ms is a pipe dream due to backhaul distance.  I think once or twice I connected with 12ms first hop.