Just looking for guidance on why the downstream synch rate is so far below the maximum attainable rate
A late reply to this, but the answer to this question is fairly simple:
DLM has intervened, and asked for interleaving and error correction to be turned on. This has three noticeable effects on the link:
- Interleaving causes additional latency, so ping times are higher. DLM's first choice of intervention usually adds 8ms, but you usually see slightly higher increases in ping times.
- Error Correction (aka FEC) steals some of your bandwidth, and re-uses it to help correct errors on the line, so you see your sync speed drop. On the statistics output, the modem shows us the sync speed *
after* the FEC overhead has already been removed. However, the "max attainable" value is calculated *
before* reduction of the FEC overhead.
- There is a second curious side-effect when FEC is turned on: The "max attainable" value actually goes up, and usually by about the same amount as the sync speed went down.
The overall impact of DLM is usually that you see the sync speed drop by about 10%, and the max attainable rise by about the same amount.
As an example, my line recently had DLM intervene for 2 months. During that time, I could see
- Sync speed dropped by around 6Mbps
- Max Attainable increased by 6Mbps
- The error-correction overhead used about 18% of the bandwidth, which adds up to about 15Mbps.
When DLM removed the intervention, the sync speed went back up by 6Mbps, and the max attainable came down by 6Mbps again.
and if the FEC error count is excessive at almost 180 million.
It sounds high, doesn't it?
When DLM intervened, and turned error-correction on, the modems respond by splitting your data down into smaller blocks, and adding protective data into the blocks (to allow the correction process to work at the other end). Because the blocks are smaller (in your case, one-thirtieth of the size), the counts can mount up quickly - so large numbers are, by themselves, not scary.
What the other figures show is this:
- The modems have transferred 1.5bn of these smaller RS blocks.
- 180m of the blocks had a fault that could be fixed. That is about 12% of the blocks.
- Just under 100k of the blocks had faults that could not be fixed. That is 0.01% of the blocks.
That 12% suggests that you really do need the error correction process to be in place. There is plenty of noise in the environment that needs to be dealt with somehow. But the 0.01% figure means that the error correction is working effectively with the current settings. In addition the graph of ES (Errored seconds) showed that you had accumulated around 150 ES's within 24 hours - that too suggests DLM will continue at this level of intervention.