Hello again,
I know it has been talked about before, and kitz.co.uk have some older details on it.
But my question is, Is there any published information available on DLM and does it really vary between ISP, i.e. BT Retail and Sky for example.
For exchange-based ADSL, it does indeed work differently across the main providers.
In BT-Wholesale, 20CN equipment tends to use the older form of DLM, known as Automatic Margin Adjustment. 21CN equipment integrates this alongside banding, more properly known as Tiered Rate Adaption. Both systems incorporate the use of FEC for error correction, which makes use of interleaving to make the error correction work more effectively.
Someone
posted a table of MTBE thresholds on the Plusnet forum that suggests, for BT's 21CN, what is being monitored - either the error rate (CRC's per 24 hours) or the rate of errored seconds per 24 hours (as MTBE, mean time between errors), plus the rate of resyncs/disconnections (as MTBR). Being red or green suggests that a line would be bumped up or down a list of profiles, which determines the next margin/band/FEC setting to use;
a further post there expands upon this nicely.
I don't know what Sky's DLM techniques are, but they certainly seem to go for a softer startup, which looks like TRA in reverse - they seem to start you in a lower speed band (and presumably measure the error rate), then turn the speed up a little after a day. Repeated a few times, if your line can cope.
Openreach FTTC still uses FEC, but has a broader variety of settings. It doesn't use AMA, but does use TRA.
The NICC published a
report on DSM techniques in the UK, dated 2010, but it doesn't name ISPs (DSM is slightly wider-reaching than, but includes, DLM). From the date, you can tell that this applies more to ADSL varieties than FTTC.
According to NICC, their
2014 to-do list includes more work on DSM, alongside exchange-based VDSL, plus use-cases for vectoring.
I'm mainly interested in DLM in relation to FTTC but any solid details on BT's / UK based implementations would be very useful!
The one thing I can add is that the VDSL2 specification includes the ability for the modem to signal loss-of-power to the DSLAM, aka "dying gasp".
Whether the DSLAM makes use of it is unknown, but if it does, then this implies it is better to power-down a modem rather than unplugging the RJ11 cable. In either case, the DLM for FTTC seems generally more immune to re-syncs caused by "user interference".
FEC (plus interleaving) seems to be the main tool used by DLM in FTTC - with the aim of making the line report very few errors. The theory seems to be that NGA-like speeds on internet access leads to users wanting more video services. Increased expectations imply that users will want a high-quality (broadcast-like) video service, and that implies very low packet loss, as they don't include a TCP layer for re-transmission.
The downside of FEC is that it steals 15-30% of your line's bandwidth (usually around 20%), while the associated interleaving adds a minimum of 8ms to latency.
There is an alternative to FEC that puts re-transmission of faulty packets down at the physical layer, rather than up at the TCP layer. This is known as PHYR. Will Openreach ever implement it? Unknown - but in the requirements for self-install FTTC modems, it is one of the requirements placed on modems.
Edit: Add FEC and PHYR for FTTC
Edited by deleted (Tue 04-Feb-14 11:17:43)
Pages in this thread: 
Print Thread
deleted
